Return-Path: jsg@goblin.cx Delivery-Date: Tue Oct 10 09:09:06 2006 Received: from min.bytemine.net (min.bytemine.net [134.106.146.209]) by cvs.openbsd.org (8.13.8/8.12.1) with ESMTP id k9AF95Hk011306 (version=TLSv1/SSLv3 cipher=DHE-DSS-AES256-SHA bits=256 verify=NO) for ; Tue, 10 Oct 2006 09:09:06 -0600 (MDT) Received: from dsl-210-15-216-215-static.vic.netspace.net.au ([210.15.216.215]:51060 helo=carla.home) by min.bytemine.net with esmtps (TLSv1:AES256-SHA:256) (Exim 4.60) (envelope-from ) id 1GXJEb-0004B2-BY for deraadt@cvs.openbsd.org; Tue, 10 Oct 2006 17:09:34 +0200 Received: from carla.home (jsg@localhost.home [127.0.0.1]) by carla.home (8.13.6/8.13.6) with ESMTP id k9AF9Qnt002664 for ; Wed, 11 Oct 2006 01:09:26 +1000 (EST) Received: (from jsg@localhost) by carla.home (8.13.6/8.13.1/Submit) id k9AF9Q1u028338 for deraadt@cvs.openbsd.org; Wed, 11 Oct 2006 01:09:26 +1000 (EST) Resent-From: jsg@goblin.cx Resent-Date: Wed, 11 Oct 2006 01:09:26 +1000 Resent-Message-ID: <20061010150926.GA3250@mail.netspace.net.au> Resent-To: Theo de Raadt Date: Fri, 15 Sep 2006 00:56:26 +1000 From: Jonathan Gray To: Dan Williams Cc: jg@laptop.org, mtosatti@redhat.com Subject: Re: Marvell 88W8388 documentation Message-ID: <20060914145626.GA8066@mail.netspace.net.au> References: <20060914124700.GA21474@mail.netspace.net.au> <1158242721.2634.26.camel@localhost.localdomain> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1158242721.2634.26.camel@localhost.localdomain> User-Agent: Mutt/1.5.11 On Thu, Sep 14, 2006 at 10:05:21AM -0400, Dan Williams wrote: > > On Thu, 2006-09-14 at 22:47 +1000, Jonathan Gray wrote: > > Hi guys, > > > > Please correct me if I am wrong but it seems that documentation > > for Marvell's 88W8388's is not publically available without > > signing an NDA? > > > > If this is the case why did a project that seems to pride > > itself of openess agree to deal with such a company? > > Drivers written under NDA tend to be full of magic numbers, > > near impossible for others to properly maintain and > > totally against the spirit of open projects. > > In the very least, the driver is open-source and usage of magic numbers > is no worse than the majority of current Linux wireless drivers, even > those developed from device specifications provided by the manufacturer. > If you look at the source, there's actually an amazing lack of magic > numbers [1], and almost all of the numbers used have meaningful names in > the header files. > > Getting specs for the chipset is expected in the near-term, but a lot of > stuff has to happen in parallel here. That's great news. > > > I really think you should push for Marvell to give out > > documentation without them forcing NDAs onto people. > > Are there specific things you're looking for that would require you to > sign an NDA with Marvell to work on? That would be useful information > to help determine where the existing opened code & resources fall short > and what bits need more transparency. A driver for their recent Ethernet chipsets, a driver for their SATA chipsets, a driver for their wireless chipsets, ongoing maintenance and de magic'ing driver for their Ethernet PHYs. You'll note that the author of the Linux driver for the former also signed an NDA. I work on the OpenBSD kernel in my spare time, we cover most of the major wireless chipsets or have documentation from vendors without signing NDAs. Marvell is one of the main sticking points and are sadly gaining a lot of marketshare over the last few years. But I really think it is in your best interests to not artifically limit the amount of people who can help you. Normally people writing drivers have little leverage with asking vendors to do anything as they aren't seen as the client. The folks over at http://www.cuwireless.net/ recently got a $500,000 US government grant to work on their NetBSD based mesh routing system, I imagine they would be quite interested in discussing some of their research and helping out, given the chance to. > > > Failing that I'm sure there are other vendors > > who would be willing to be more helpful. > > Possibly, but Marvell had the chip, was quite willing to do the work to > modify the firmware bits to OLPC specifications, was willing to > open-source their driver, etc. Other vendors weren't nearly as willing > as Marvell was to help the project out, so I'm not so sure your > assertion here is correct. I suspect you really wanted to talk to were Zydas, they had the usb wireless market and were highly regarded. Now they have been aquired by Atheros it is unlikely they'll be allowed to release documentation, source code and allow firmware redistribution like they have in the past. Ralink have been very helpful with questions, documentation, access to engineers and the like in the past. I suspect they would at least be interested to hear of the requirements. Realtek seem to want to help or give you the run around depending on who you talk to. They could perhaps be another option though. VIA have pretty much ignored requests for documentation but perhaps have some interesting technology. I guess you really need something with an ARM/MIPS chip if you want it to route while the main processor is suspended. Which is rather sad, as that puts things into the arena of the large companies that implement most of their stack in the micro and tend to cost more. > > Dan > > [1] compared to, say, the airo driver which was even developed with some > help from Cisco I have tried to find documentation on Aironet devices as well, nothing is out there, again if anyone knows otherwise, I'm all ears. Also I suspect it was Aironet before Cisco purchased them that would have helped. Jonathan Return-Path: jsg@goblin.cx Delivery-Date: Tue Oct 10 09:09:21 2006 Received: from min.bytemine.net (min.bytemine.net [134.106.146.209]) by cvs.openbsd.org (8.13.8/8.12.1) with ESMTP id k9AF9Jpb010528 (version=TLSv1/SSLv3 cipher=DHE-DSS-AES256-SHA bits=256 verify=NO) for ; Tue, 10 Oct 2006 09:09:20 -0600 (MDT) Received: from dsl-210-15-216-215-static.vic.netspace.net.au ([210.15.216.215]:63158 helo=carla.home) by min.bytemine.net with esmtps (TLSv1:AES256-SHA:256) (Exim 4.60) (envelope-from ) id 1GXJEq-00059f-Eh for deraadt@cvs.openbsd.org; Tue, 10 Oct 2006 17:09:49 +0200 Received: from carla.home (jsg@localhost.home [127.0.0.1]) by carla.home (8.13.6/8.13.6) with ESMTP id k9AF9iaf010879 for ; Wed, 11 Oct 2006 01:09:44 +1000 (EST) Received: (from jsg@localhost) by carla.home (8.13.6/8.13.1/Submit) id k9AF9iH8011091 for deraadt@cvs.openbsd.org; Wed, 11 Oct 2006 01:09:44 +1000 (EST) Resent-From: jsg@goblin.cx Resent-Date: Wed, 11 Oct 2006 01:09:44 +1000 Resent-Message-ID: <20061010150944.GB3250@mail.netspace.net.au> Resent-To: Theo de Raadt Date: Thu, 14 Sep 2006 22:47:00 +1000 From: Jonathan Gray To: dcbw@redhat.com, jg@laptop.org, mtosatti@redhat.com Subject: Marvell 88W8388 documentation Message-ID: <20060914124700.GA21474@mail.netspace.net.au> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.5.11 Hi guys, Please correct me if I am wrong but it seems that documentation for Marvell's 88W8388's is not publically available without signing an NDA? If this is the case why did a project that seems to pride itself of openess agree to deal with such a company? Drivers written under NDA tend to be full of magic numbers, near impossible for others to properly maintain and totally against the spirit of open projects. I really think you should push for Marvell to give out documentation without them forcing NDAs onto people. Failing that I'm sure there are other vendors who would be willing to be more helpful. Jonathan Return-Path: jsg@goblin.cx Delivery-Date: Tue Oct 10 09:10:26 2006 Received: from min.bytemine.net (min.bytemine.net [134.106.146.209]) by cvs.openbsd.org (8.13.8/8.12.1) with ESMTP id k9AFAOAw022282 (version=TLSv1/SSLv3 cipher=DHE-DSS-AES256-SHA bits=256 verify=NO) for ; Tue, 10 Oct 2006 09:10:25 -0600 (MDT) Received: from dsl-210-15-216-215-static.vic.netspace.net.au ([210.15.216.215]:60027 helo=carla.home) by min.bytemine.net with esmtps (TLSv1:AES256-SHA:256) (Exim 4.60) (envelope-from ) id 1GXJFs-0004c7-Gw for deraadt@cvs.openbsd.org; Tue, 10 Oct 2006 17:10:53 +0200 Received: from carla.home (jsg@localhost.home [127.0.0.1]) by carla.home (8.13.6/8.13.6) with ESMTP id k9AFAmeA023501 for ; Wed, 11 Oct 2006 01:10:48 +1000 (EST) Received: (from jsg@localhost) by carla.home (8.13.6/8.13.1/Submit) id k9AFAmmJ026343 for deraadt@cvs.openbsd.org; Wed, 11 Oct 2006 01:10:48 +1000 (EST) Resent-From: jsg@goblin.cx Resent-Date: Wed, 11 Oct 2006 01:10:47 +1000 Resent-Message-ID: <20061010151047.GC3250@mail.netspace.net.au> Resent-To: Theo de Raadt Delivered-To: unknown Received: from mail.goblin.cx (69.80.208.30) by carla.home with POP3; 14 Sep 2006 14:16:41 -0000 Received: from alligate2.nshosts.com (unverified [10.0.1.44]) by mail8.nshosts.com (Vircom SMTPRS 5.3.232) with ESMTP id for ; Thu, 14 Sep 2006 08:03:37 -0600 Received: from mx1.redhat.com [66.187.233.31] by alligate2.nshosts.com (Alligate(TM) SMTP Gateway v2.6.5.25) with ESMPT id for ; Thu, 14 Sep 2006 08:03:28 -0700 Received: from int-mx1.corp.redhat.com (int-mx1.corp.redhat.com [172.16.52.254]) by mx1.redhat.com (8.12.11.20060308/8.12.11) with ESMTP id k8EE3VH8002751; Thu, 14 Sep 2006 10:03:31 -0400 Received: from mail.boston.redhat.com (mail.boston.redhat.com [172.16.76.12]) by int-mx1.corp.redhat.com (8.12.11.20060308/8.12.11) with ESMTP id k8EE3VfQ008843; Thu, 14 Sep 2006 10:03:31 -0400 Received: from [10.13.248.15] (vpn-248-15.boston.redhat.com [10.13.248.15]) by mail.boston.redhat.com (8.12.8/8.12.8) with ESMTP id k8EE3TY9028201; Thu, 14 Sep 2006 10:03:30 -0400 Subject: Re: Marvell 88W8388 documentation From: Dan Williams To: Jonathan Gray Cc: jg@laptop.org, mtosatti@redhat.com In-Reply-To: <20060914124700.GA21474@mail.netspace.net.au> References: <20060914124700.GA21474@mail.netspace.net.au> Content-Type: text/plain Date: Thu, 14 Sep 2006 10:05:21 -0400 Message-Id: <1158242721.2634.26.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Evolution 2.8.0 (2.8.0-1.fc6) Content-Transfer-Encoding: 7bit X-MXRate-Prob: 0 X-MXRate-Country: US X-MXRate-Action: NONE X-Alligate-Grey: Skipped X-Alligate-In: Passed - Adult: 0 (Req: 100) Spam: 16 (Req: 100) Tot: 16 (Req: 100) X-Alligate-QueueFile: 010974219.dta X-Alligate-XFrom: [66.187.233.31] United States (US) X-Alligate-XTo: (jsg@goblin.cx) On Thu, 2006-09-14 at 22:47 +1000, Jonathan Gray wrote: > Hi guys, > > Please correct me if I am wrong but it seems that documentation > for Marvell's 88W8388's is not publically available without > signing an NDA? > > If this is the case why did a project that seems to pride > itself of openess agree to deal with such a company? > Drivers written under NDA tend to be full of magic numbers, > near impossible for others to properly maintain and > totally against the spirit of open projects. In the very least, the driver is open-source and usage of magic numbers is no worse than the majority of current Linux wireless drivers, even those developed from device specifications provided by the manufacturer. If you look at the source, there's actually an amazing lack of magic numbers [1], and almost all of the numbers used have meaningful names in the header files. Getting specs for the chipset is expected in the near-term, but a lot of stuff has to happen in parallel here. > I really think you should push for Marvell to give out > documentation without them forcing NDAs onto people. Are there specific things you're looking for that would require you to sign an NDA with Marvell to work on? That would be useful information to help determine where the existing opened code & resources fall short and what bits need more transparency. > Failing that I'm sure there are other vendors > who would be willing to be more helpful. Possibly, but Marvell had the chip, was quite willing to do the work to modify the firmware bits to OLPC specifications, was willing to open-source their driver, etc. Other vendors weren't nearly as willing as Marvell was to help the project out, so I'm not so sure your assertion here is correct. Dan [1] compared to, say, the airo driver which was even developed with some help from Cisco Return-Path: jsg@goblin.cx Delivery-Date: Tue Oct 10 09:10:48 2006 Received: from min.bytemine.net (min.bytemine.net [134.106.146.209]) by cvs.openbsd.org (8.13.8/8.12.1) with ESMTP id k9AFAlCA010053 (version=TLSv1/SSLv3 cipher=DHE-DSS-AES256-SHA bits=256 verify=NO) for ; Tue, 10 Oct 2006 09:10:48 -0600 (MDT) Received: from dsl-210-15-216-215-static.vic.netspace.net.au ([210.15.216.215]:51287 helo=carla.home) by min.bytemine.net with esmtps (TLSv1:AES256-SHA:256) (Exim 4.60) (envelope-from ) id 1GXJGF-0008BL-Ah for deraadt@cvs.openbsd.org; Tue, 10 Oct 2006 17:11:16 +0200 Received: from carla.home (jsg@localhost.home [127.0.0.1]) by carla.home (8.13.6/8.13.6) with ESMTP id k9AFBBob026830 for ; Wed, 11 Oct 2006 01:11:11 +1000 (EST) Received: (from jsg@localhost) by carla.home (8.13.6/8.13.1/Submit) id k9AFBBZS032436 for deraadt@cvs.openbsd.org; Wed, 11 Oct 2006 01:11:11 +1000 (EST) Resent-From: jsg@goblin.cx Resent-Date: Wed, 11 Oct 2006 01:11:11 +1000 Resent-Message-ID: <20061010151111.GD3250@mail.netspace.net.au> Resent-To: Theo de Raadt Delivered-To: unknown Received: from mail.goblin.cx (69.80.208.30) by carla.home with POP3; 14 Sep 2006 16:42:17 -0000 Received: from alligate6.nshosts.com (unverified [10.0.1.46]) by mail3.nshosts.com (Vircom SMTPRS 5.3.232) with ESMTP id for ; Thu, 14 Sep 2006 10:27:13 -0600 Received: from mx1.redhat.com [66.187.233.31] by alligate6.nshosts.com (Alligate(TM) SMTP Gateway v2.6.5.25) with ESMPT id <947057C36B6B2C89.8E6522546AE2DE08@alligate6.nshosts.com> for ; Thu, 14 Sep 2006 10:23:57 -0700 Received: from int-mx1.corp.redhat.com (int-mx1.corp.redhat.com [172.16.52.254]) by mx1.redhat.com (8.12.11.20060308/8.12.11) with ESMTP id k8EGNxmj030510; Thu, 14 Sep 2006 12:23:59 -0400 Received: from mail.boston.redhat.com (mail.boston.redhat.com [172.16.76.12]) by int-mx1.corp.redhat.com (8.12.11.20060308/8.12.11) with ESMTP id k8EGNwEt002086; Thu, 14 Sep 2006 12:23:58 -0400 Received: from [10.13.248.15] (vpn-248-15.boston.redhat.com [10.13.248.15]) by mail.boston.redhat.com (8.12.8/8.12.8) with ESMTP id k8EGNrY9014449; Thu, 14 Sep 2006 12:23:56 -0400 Subject: Re: Marvell 88W8388 documentation From: Dan Williams To: Jonathan Gray Cc: jg@laptop.org, mtosatti@redhat.com In-Reply-To: <20060914145626.GA8066@mail.netspace.net.au> References: <20060914124700.GA21474@mail.netspace.net.au> <1158242721.2634.26.camel@localhost.localdomain> <20060914145626.GA8066@mail.netspace.net.au> Content-Type: text/plain Date: Thu, 14 Sep 2006 12:25:44 -0400 Message-Id: <1158251144.4675.4.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Evolution 2.8.0 (2.8.0-1.fc6) Content-Transfer-Encoding: 7bit X-MXRate-Prob: 0 X-MXRate-Country: US X-MXRate-Action: NONE X-Alligate-Grey: Skipped X-Alligate-In: Passed - Adult: 0 (Req: 100) Spam: 18 (Req: 100) Tot: 18 (Req: 100) X-Alligate-QueueFile: 010315403.dta X-Alligate-XFrom: [66.187.233.31] United States (US) X-Alligate-XTo: (jsg@goblin.cx) On Fri, 2006-09-15 at 00:56 +1000, Jonathan Gray wrote: > On Thu, Sep 14, 2006 at 10:05:21AM -0400, Dan Williams wrote: > > > > On Thu, 2006-09-14 at 22:47 +1000, Jonathan Gray wrote: > > > Hi guys, > > > > > > Please correct me if I am wrong but it seems that documentation > > > for Marvell's 88W8388's is not publically available without > > > signing an NDA? > > > > > > If this is the case why did a project that seems to pride > > > itself of openess agree to deal with such a company? > > > Drivers written under NDA tend to be full of magic numbers, > > > near impossible for others to properly maintain and > > > totally against the spirit of open projects. > > > > In the very least, the driver is open-source and usage of magic numbers > > is no worse than the majority of current Linux wireless drivers, even > > those developed from device specifications provided by the manufacturer. > > If you look at the source, there's actually an amazing lack of magic > > numbers [1], and almost all of the numbers used have meaningful names in > > the header files. > > > > Getting specs for the chipset is expected in the near-term, but a lot of > > stuff has to happen in parallel here. > > That's great news. > > > > > > I really think you should push for Marvell to give out > > > documentation without them forcing NDAs onto people. > > > > Are there specific things you're looking for that would require you to > > sign an NDA with Marvell to work on? That would be useful information > > to help determine where the existing opened code & resources fall short > > and what bits need more transparency. > > A driver for their recent Ethernet chipsets, a driver for their SATA > chipsets, a driver for their wireless chipsets, ongoing maintenance > and de magic'ing driver for their Ethernet PHYs. You'll note > that the author of the Linux driver for the former also signed an NDA. Right; that would be nice but everything in that list besides the wireless driver for the 8388 is beyond our scope. Hopefully though, the process of opening up the driver and specs for the 8388, which we have so far held Marvell's hand through, will show them that it's not painful and that they have nothing to fear from opening up their stuff. But that's all we can do now, and all we have leverage over. > I work on the OpenBSD kernel in my spare time, we cover most of > the major wireless chipsets or have documentation from vendors without > signing NDAs. Marvell is one of the main sticking points > and are sadly gaining a lot of marketshare over the last > few years. > > But I really think it is in your best interests to not > artifically limit the amount of people who can help you. > Normally people writing drivers have little leverage > with asking vendors to do anything as they aren't seen > as the client. > > The folks over at http://www.cuwireless.net/ recently got > a $500,000 US government grant to work on their NetBSD > based mesh routing system, I imagine they would be > quite interested in discussing some of their research > and helping out, given the chance to. > > > > > > Failing that I'm sure there are other vendors > > > who would be willing to be more helpful. > > > > Possibly, but Marvell had the chip, was quite willing to do the work to > > modify the firmware bits to OLPC specifications, was willing to > > open-source their driver, etc. Other vendors weren't nearly as willing > > as Marvell was to help the project out, so I'm not so sure your > > assertion here is correct. > > I suspect you really wanted to talk to were Zydas, they had > the usb wireless market and were highly regarded. Now they have > been aquired by Atheros it is unlikely they'll be allowed to > release documentation, source code and allow firmware redistribution > like they have in the past. > > Ralink have been very helpful with questions, documentation, access > to engineers and the like in the past. I suspect they would at least > be interested to hear of the requirements. > > Realtek seem to want to help or give you the run around depending > on who you talk to. They could perhaps be another option though. > > VIA have pretty much ignored requests for documentation but > perhaps have some interesting technology. > > I guess you really need something with an ARM/MIPS chip > if you want it to route while the main processor is suspended. > Which is rather sad, as that puts things into the arena > of the large companies that implement most of their stack > in the micro and tend to cost more. > > > > > Dan > > > > [1] compared to, say, the airo driver which was even developed with some > > help from Cisco > > I have tried to find documentation on Aironet devices as well, nothing > is out there, again if anyone knows otherwise, I'm all ears. Also I > suspect it was Aironet before Cisco purchased them that would have helped. There isn't any, and I've tried to get some through Red Hat channels, even Cisco's IT department wants us to fix up their own driver, but they have decided the parts are obsolete and do not wish to spend resources even sending us a PDF of the "Aironet 4500/4800 Developer's Manual" which is referenced at the top. Dan Return-Path: jsg@goblin.cx Delivery-Date: Tue Oct 10 09:11:02 2006 Received: from min.bytemine.net (min.bytemine.net [134.106.146.209]) by cvs.openbsd.org (8.13.8/8.12.1) with ESMTP id k9AFB0ZQ000255 (version=TLSv1/SSLv3 cipher=DHE-DSS-AES256-SHA bits=256 verify=NO) for ; Tue, 10 Oct 2006 09:11:02 -0600 (MDT) Received: from dsl-210-15-216-215-static.vic.netspace.net.au ([210.15.216.215]:59947 helo=carla.home) by min.bytemine.net with esmtps (TLSv1:AES256-SHA:256) (Exim 4.60) (envelope-from ) id 1GXJGS-0006a7-UZ for deraadt@cvs.openbsd.org; Tue, 10 Oct 2006 17:11:30 +0200 Received: from carla.home (jsg@localhost.home [127.0.0.1]) by carla.home (8.13.6/8.13.6) with ESMTP id k9AFBO3C016938 for ; Wed, 11 Oct 2006 01:11:24 +1000 (EST) Received: (from jsg@localhost) by carla.home (8.13.6/8.13.1/Submit) id k9AFBOFM001736 for deraadt@cvs.openbsd.org; Wed, 11 Oct 2006 01:11:24 +1000 (EST) Resent-From: jsg@goblin.cx Resent-Date: Wed, 11 Oct 2006 01:11:24 +1000 Resent-Message-ID: <20061010151124.GE3250@mail.netspace.net.au> Resent-To: Theo de Raadt Delivered-To: unknown Received: from mail.goblin.cx (69.80.208.30) by carla.home with POP3; 14 Sep 2006 15:32:56 -0000 Received: from alligate5.nshosts.com (unverified [10.0.1.45]) by MAIL9.nshosts.com (Vircom SMTPRS 5.3.232) with ESMTP id for ; Thu, 14 Sep 2006 09:12:44 -0600 Received: from alnrmhc14.comcast.net [206.18.177.54] by alligate5.nshosts.com (Alligate(TM) SMTP Gateway v2.6.5.25) with ESMPT id for ; Thu, 14 Sep 2006 09:15:48 -0700 Received: from wireless-19-34.media.mit.edu ([18.85.19.34]) by comcast.net (alnrmhc14) with SMTP id <20060914151554b1400m146fe>; Thu, 14 Sep 2006 15:15:54 +0000 Subject: Re: Marvell 88W8388 documentation From: Jim Gettys Reply-To: jg@laptop.org To: Jonathan Gray Cc: Dan Williams , mtosatti@redhat.com In-Reply-To: <20060914145626.GA8066@mail.netspace.net.au> References: <20060914124700.GA21474@mail.netspace.net.au> <1158242721.2634.26.camel@localhost.localdomain> <20060914145626.GA8066@mail.netspace.net.au> Content-Type: text/plain Organization: OLPC Date: Thu, 14 Sep 2006 11:15:53 -0400 Message-Id: <1158246953.5871.215.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Evolution 2.6.1 Content-Transfer-Encoding: 7bit X-MXRate-Prob: -1 X-MXRate-Country: US X-MXRate-Action: ALLOW X-Alligate-Grey: Skipped X-Alligate-In: Passed - Adult: 0 (Req: 100) Spam: 0 (Req: 100) Tot: 0 (Req: 100) X-Alligate-QueueFile: 020028166.dta X-Alligate-XFrom: [206.18.177.54] United States (US) X-Alligate-XTo: (jsg@goblin.cx) On Fri, 2006-09-15 at 00:56 +1000, Jonathan Gray wrote: > > A driver for their recent Ethernet chipsets, a driver for their SATA > chipsets, a driver for their wireless chipsets, ongoing maintenance > and de magic'ing driver for their Ethernet PHYs. You'll note > that the author of the Linux driver for the former also signed an NDA. > > I work on the OpenBSD kernel in my spare time, we cover most of > the major wireless chipsets or have documentation from vendors without > signing NDAs. Marvell is one of the main sticking points > and are sadly gaining a lot of marketshare over the last > few years. > > But I really think it is in your best interests to not > artifically limit the amount of people who can help you. > Normally people writing drivers have little leverage > with asking vendors to do anything as they aren't seen > as the client. We're very aware of this. That's why we're working on the issues around this. I just wanted you to understand that there were no alternatives, and that Marvell could not give us all we want; it wasn't under their control. > > The folks over at http://www.cuwireless.net/ recently got > a $500,000 US government grant to work on their NetBSD > based mesh routing system, I imagine they would be > quite interested in discussing some of their research > and helping out, given the chance to. > I think they have been in touch with Michailis. > > > > > Failing that I'm sure there are other vendors > > > who would be willing to be more helpful. > > > > Possibly, but Marvell had the chip, was quite willing to do the work to > > modify the firmware bits to OLPC specifications, was willing to > > open-source their driver, etc. Other vendors weren't nearly as willing > > as Marvell was to help the project out, so I'm not so sure your > > assertion here is correct. > > I suspect you really wanted to talk to were Zydas, they had > the usb wireless market and were highly regarded. Now they have > been aquired by Atheros it is unlikely they'll be allowed to > release documentation, source code and allow firmware redistribution > like they have in the past. > > Ralink have been very helpful with questions, documentation, access > to engineers and the like in the past. I suspect they would at least > be interested to hear of the requirements. > > Realtek seem to want to help or give you the run around depending > on who you talk to. They could perhaps be another option though. > > VIA have pretty much ignored requests for documentation but > perhaps have some interesting technology. As of this spring, *no* other vendor had the capability to allow the mesh code to run with the main processor off; Michails was/is very in touch with what the different vendors are doing. Athleros was unwilling to commit to being able to do this even in their next generation (which would have been to late for us). That the Marvell chip *is* capable enough to run autonomously was a great surprise; Marvell has previously sold only to the embedded market, and has not competed for desktop/laptop business before this. We let them all know what we need when they ask. And USB is *not* desirable; it just happens to be the only bus available on this generation part in common with the Geode. Future parts will be interfaced via different buses. > > I guess you really need something with an ARM/MIPS chip > if you want it to route while the main processor is suspended. Yes, this is *absolutely essential* to us. It is the difference between 300mw power consumption and 2-3 watts power consumption. If we were in the 2 watt range (minimum power with the main CPU on, etc.), the kids would have to disable their machines from the mesh all the time to conserve battery. A small kid can only generate 7-10 watts. So, from one point of view, this capability is one of the 3 things that make this machine truly novel and desirable. > Which is rather sad, as that puts things into the arena > of the large companies that implement most of their stack > in the micro and tend to cost more. > Actually, Marvell was competitive on prices as well, with the less capable chips (e.g. Athleros). I can't tell you what price we got. Regards, - Jim > > > > Dan > > > > [1] compared to, say, the airo driver which was even developed with some > > help from Cisco > > I have tried to find documentation on Aironet devices as well, nothing > is out there, again if anyone knows otherwise, I'm all ears. Also I > suspect it was Aironet before Cisco purchased them that would have helped. > > Jonathan -- Jim Gettys One Laptop Per Child Return-Path: jsg@goblin.cx Delivery-Date: Tue Oct 10 09:11:23 2006 Received: from min.bytemine.net (min.bytemine.net [134.106.146.209]) by cvs.openbsd.org (8.13.8/8.12.1) with ESMTP id k9AFBMg5028682 (version=TLSv1/SSLv3 cipher=DHE-DSS-AES256-SHA bits=256 verify=NO) for ; Tue, 10 Oct 2006 09:11:23 -0600 (MDT) Received: from dsl-210-15-216-215-static.vic.netspace.net.au ([210.15.216.215]:61703 helo=carla.home) by min.bytemine.net with esmtps (TLSv1:AES256-SHA:256) (Exim 4.60) (envelope-from ) id 1GXJGo-0006RR-D2 for deraadt@cvs.openbsd.org; Tue, 10 Oct 2006 17:11:51 +0200 Received: from carla.home (jsg@localhost.home [127.0.0.1]) by carla.home (8.13.6/8.13.6) with ESMTP id k9AFBhMq008885 for ; Wed, 11 Oct 2006 01:11:43 +1000 (EST) Received: (from jsg@localhost) by carla.home (8.13.6/8.13.1/Submit) id k9AFBhO1028197 for deraadt@cvs.openbsd.org; Wed, 11 Oct 2006 01:11:43 +1000 (EST) Resent-From: jsg@goblin.cx Resent-Date: Wed, 11 Oct 2006 01:11:43 +1000 Resent-Message-ID: <20061010151143.GF3250@mail.netspace.net.au> Resent-To: Theo de Raadt Delivered-To: unknown Received: from mail.goblin.cx (69.80.208.30) by carla.home with POP3; 14 Sep 2006 14:56:49 -0000 Received: from alligate5.nshosts.com (unverified [10.0.1.45]) by MAIL9.nshosts.com (Vircom SMTPRS 5.3.232) with ESMTP id for ; Thu, 14 Sep 2006 08:13:32 -0600 Received: from alnrmhc12.comcast.net [204.127.225.92] by alligate5.nshosts.com (Alligate(TM) SMTP Gateway v2.6.5.25) with ESMPT id for ; Thu, 14 Sep 2006 08:16:26 -0700 Received: from wireless-19-34.media.mit.edu ([18.85.19.34]) by comcast.net (alnrmhc12) with SMTP id <20060914141632b1200pcvppe>; Thu, 14 Sep 2006 14:16:32 +0000 Subject: Re: Marvell 88W8388 documentation From: Jim Gettys Reply-To: jg@laptop.org To: Dan Williams Cc: Jonathan Gray , mtosatti@redhat.com In-Reply-To: <1158242721.2634.26.camel@localhost.localdomain> References: <20060914124700.GA21474@mail.netspace.net.au> <1158242721.2634.26.camel@localhost.localdomain> Content-Type: text/plain Organization: OLPC Date: Thu, 14 Sep 2006 10:16:30 -0400 Message-Id: <1158243391.5871.196.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Evolution 2.6.1 Content-Transfer-Encoding: 7bit X-MXRate-Prob: -1 X-MXRate-Country: US X-MXRate-Action: ALLOW X-Alligate-Grey: Skipped X-Alligate-In: Passed - Adult: 0 (Req: 100) Spam: 0 (Req: 100) Tot: 0 (Req: 100) X-Alligate-QueueFile: 020021598.dta X-Alligate-XFrom: [204.127.225.92] United States (US) X-Alligate-XTo: (jsg@goblin.cx) On Thu, 2006-09-14 at 10:05 -0400, Dan Williams wrote: > > Failing that I'm sure there are other vendors > > who would be willing to be more helpful. > > Possibly, but Marvell had the chip, was quite willing to do the work to > modify the firmware bits to OLPC specifications, was willing to > open-source their driver, etc. Other vendors weren't nearly as willing > as Marvell was to help the project out, so I'm not so sure your > assertion here is correct. > It is moot: no other chips had the capability we needed present in the Marvell chip. The other vendors (which in general are much less helpful to free software) were not even willing to consider adding the capability in their next generation, and were suffering a severe case of "hubris", that we could not possibly go with someone else's chip. - Jim -- Jim Gettys One Laptop Per Child Replied: Wed, 04 Oct 2006 12:38:34 -0600 Replied: "Jonathan Gray dcbw@redhat.com, jg@laptop.org, mtosatti@redhat.com" Return-Path: jsg@goblin.cx Delivery-Date: Thu Sep 14 06:47:19 2006 Received: from min.bytemine.net (min.bytemine.net [134.106.146.209]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k8EClHnM017171 (version=TLSv1/SSLv3 cipher=DHE-DSS-AES256-SHA bits=256 verify=NO) for ; Thu, 14 Sep 2006 06:47:19 -0600 (MDT) Received: from dsl-210-15-216-215-static.vic.netspace.net.au ([210.15.216.215]:61248 helo=carla.home) by min.bytemine.net with esmtps (TLSv1:AES256-SHA:256) (Exim 4.60) (envelope-from ) id 1GNqcU-0003vW-Je; Thu, 14 Sep 2006 14:47:08 +0200 Received: from carla.home (jsg@localhost.home [127.0.0.1]) by carla.home (8.13.6/8.13.6) with ESMTP id k8ECl10M027118; Thu, 14 Sep 2006 22:47:01 +1000 (EST) Received: (from jsg@localhost) by carla.home (8.13.6/8.13.1/Submit) id k8ECl0Qn006521; Thu, 14 Sep 2006 22:47:00 +1000 (EST) Date: Thu, 14 Sep 2006 22:47:00 +1000 From: Jonathan Gray To: dcbw@redhat.com, jg@laptop.org, mtosatti@redhat.com Subject: Marvell 88W8388 documentation Message-ID: <20060914124700.GA21474@mail.netspace.net.au> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.5.11 Hi guys, Please correct me if I am wrong but it seems that documentation for Marvell's 88W8388's is not publically available without signing an NDA? If this is the case why did a project that seems to pride itself of openess agree to deal with such a company? Drivers written under NDA tend to be full of magic numbers, near impossible for others to properly maintain and totally against the spirit of open projects. I really think you should push for Marvell to give out documentation without them forcing NDAs onto people. Failing that I'm sure there are other vendors who would be willing to be more helpful. Jonathan Return-Path: deraadt@cvs.openbsd.org Delivery-Date: Wed Oct 4 12:38:36 2006 Received: from cvs.openbsd.org (localhost [127.0.0.1]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k94IcY6j006189; Wed, 4 Oct 2006 12:38:34 -0600 (MDT) Message-Id: <200610041838.k94IcY6j006189@cvs.openbsd.org> To: Jonathan Gray cc: dcbw@redhat.com, jg@laptop.org, mtosatti@redhat.com cc: rms@gnu.org cc: deraadt Subject: Re: Marvell 88W8388 documentation In-reply-to: Your message of "Thu, 14 Sep 2006 22:47:00 +1000." <20060914124700.GA21474@mail.netspace.net.au> Date: Wed, 04 Oct 2006 12:38:34 -0600 From: Theo de Raadt > Please correct me if I am wrong but it seems that documentation > for Marvell's 88W8388's is not publically available without > signing an NDA? > > If this is the case why did a project that seems to pride > itself of openess agree to deal with such a company? > Drivers written under NDA tend to be full of magic numbers, > near impossible for others to properly maintain and > totally against the spirit of open projects. > > I really think you should push for Marvell to give out > documentation without them forcing NDAs onto people. > Failing that I'm sure there are other vendors > who would be willing to be more helpful. Jonathan showed me this mail he sent you about your NDA "cooperation" with Marvell for the wireless chip that you want to use for the OLPC project, so that Marvell will write you special hacks to do low-power mesh networking while the main cpu is powered off. This does not gaurantee Marvell is going to be open and release documentation for their chips though. When large players like you make such private agreements with such secretive vendors, you work against our common goals of getting more open documentation for devices. It is only with open documentation that OS groups can increase device support, and later -- keep the device drivers reliable after the device is EOL'd by the vendor. I've heard claims that you (OLPC members, Red Hat employees) think this relationship with Marvell will eventually prompt/teach them to be more open in time. Do you not realize how much of a DELUSION the history of free/open operating systems shows that point of view to be? Very few chip vendors have ever opened up unless they were pushed, let alone Marvell (who I am led to believe also has NDA's with Red Hat employees for the Marvell Yukon/Yukon 2 gigabit ethernet chips -- again one of the few closed chips). It is clear that your choices are not about opening up Marvell, but simply commercially expedient and hurtful to our common cause. You came to Marvell with potential sales of millions of units, and then completely wimped out in demanding ideals that you say you share with the community. Now other companies like Intel, Broadcom, and TI can say to us "Why should we open up, Marvell did not have to". So I must say I am extremely dissapointed you have chosen to work against the very obvious goals of "open", and I hope that in time you are made to feel ashamed of the choice you have made. Replied: Wed, 04 Oct 2006 13:41:36 -0600 Replied: "jg@laptop.org Jonathan Gray , dcbw@redhat.com, mtosatti@redhat.com, Replied: rms@gnu.org" Return-Path: jg@laptop.org Delivery-Date: Wed Oct 4 13:10:31 2006 Received: from alnrmhc14.comcast.net (alnrmhc14.comcast.net [204.127.225.94]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k94JAUJx023273 for ; Wed, 4 Oct 2006 13:10:31 -0600 (MDT) Received: from wireless-81.media.mit.edu ([18.85.18.81]) by comcast.net (alnrmhc14) with SMTP id <20061004191003b1400kgl8ne>; Wed, 4 Oct 2006 19:10:24 +0000 Subject: Re: Marvell 88W8388 documentation From: Jim Gettys Reply-To: jg@laptop.org To: Theo de Raadt Cc: Jonathan Gray , dcbw@redhat.com, mtosatti@redhat.com, rms@gnu.org In-Reply-To: <200610041838.k94IcY6j006189@cvs.openbsd.org> References: <200610041838.k94IcY6j006189@cvs.openbsd.org> Content-Type: text/plain Organization: OLPC Date: Wed, 04 Oct 2006 15:10:02 -0400 Message-Id: <1159989002.6029.710.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Evolution 2.6.1 Content-Transfer-Encoding: 7bit Theo, You may not believe this, but we are continuing to work on Marvell (push, is the term you use); it has not fallen off our list at all and is continuing to be worked by us vigorously. And there is a reality that a piece of this is out of Marvell's short term control: their firmware was based on a commercial embedded OS, not theirs to give away. There may be long term solutions here by replacing all such code, but no short term ones. We could not ask Marvell for something they could not give us. You also need to understand that this component was/is literally the only game in town to achieve our goals: *there are no other alternatives*. We had no choice at all. Not Intel. Not Broadcom. Not Athleros. That Marvell chip is currently a unique part, *and they know it*. Negotiating positions are much stronger when you have alternatives, and we did not. For just about everything else, we were able to negotiate from strength; but not for this component. Best regards, - Jim Gettys On Wed, 2006-10-04 at 12:38 -0600, Theo de Raadt wrote: > > Please correct me if I am wrong but it seems that documentation > > for Marvell's 88W8388's is not publically available without > > signing an NDA? > > > > If this is the case why did a project that seems to pride > > itself of openess agree to deal with such a company? > > Drivers written under NDA tend to be full of magic numbers, > > near impossible for others to properly maintain and > > totally against the spirit of open projects. > > > > I really think you should push for Marvell to give out > > documentation without them forcing NDAs onto people. > > Failing that I'm sure there are other vendors > > who would be willing to be more helpful. > > Jonathan showed me this mail he sent you about your NDA "cooperation" > with Marvell for the wireless chip that you want to use for the OLPC > project, so that Marvell will write you special hacks to do low-power > mesh networking while the main cpu is powered off. This does not > gaurantee Marvell is going to be open and release documentation for > their chips though. > > When large players like you make such private agreements with such > secretive vendors, you work against our common goals of getting more > open documentation for devices. It is only with open documentation > that OS groups can increase device support, and later -- keep the > device drivers reliable after the device is EOL'd by the vendor. > > I've heard claims that you (OLPC members, Red Hat employees) think > this relationship with Marvell will eventually prompt/teach them to be > more open in time. Do you not realize how much of a DELUSION the > history of free/open operating systems shows that point of view to be? > Very few chip vendors have ever opened up unless they were pushed, let > alone Marvell (who I am led to believe also has NDA's with Red Hat > employees for the Marvell Yukon/Yukon 2 gigabit ethernet chips -- > again one of the few closed chips). > > It is clear that your choices are not about opening up Marvell, but > simply commercially expedient and hurtful to our common cause. You > came to Marvell with potential sales of millions of units, and then > completely wimped out in demanding ideals that you say you share with > the community. Now other companies like Intel, Broadcom, and TI can > say to us "Why should we open up, Marvell did not have to". > > So I must say I am extremely dissapointed you have chosen to work > against the very obvious goals of "open", and I hope that in time you > are made to feel ashamed of the choice you have made. -- Jim Gettys One Laptop Per Child Replied: Thu, 05 Oct 2006 16:16:32 -0600 Replied: "jg@laptop.org Jonathan Gray , dcbw@redhat.com, mtosatti@redhat.com, Replied: rms@gnu.org" Return-Path: jg@laptop.org Delivery-Date: Wed Oct 4 14:20:48 2006 Received: from alnrmhc14.comcast.net (alnrmhc14.comcast.net [206.18.177.54]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k94KKmIe026282 for ; Wed, 4 Oct 2006 14:20:48 -0600 (MDT) Received: from wireless-81.media.mit.edu ([18.85.18.81]) by comcast.net (alnrmhc14) with SMTP id <20061004202040b1400kgp3qe>; Wed, 4 Oct 2006 20:20:42 +0000 Subject: Re: Marvell 88W8388 documentation From: Jim Gettys Reply-To: jg@laptop.org To: Theo de Raadt Cc: Jonathan Gray , dcbw@redhat.com, mtosatti@redhat.com, rms@gnu.org In-Reply-To: <200610041941.k94JfaPt019572@cvs.openbsd.org> References: <200610041941.k94JfaPt019572@cvs.openbsd.org> Content-Type: text/plain Organization: OLPC Date: Wed, 04 Oct 2006 16:20:36 -0400 Message-Id: <1159993236.6029.758.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Evolution 2.6.1 Content-Transfer-Encoding: 7bit On Wed, 2006-10-04 at 13:41 -0600, Theo de Raadt wrote: > > You may not believe this, but we are continuing to work on Marvell > > (push, is the term you use); it has not fallen off our list at all and > > is continuing to be worked by us vigorously. > > > > And there is a reality that a piece of this is out of Marvell's short > > term control: their firmware was based on a commercial embedded OS, not > > theirs to give away. There may be long term solutions here by replacing > > all such code, but no short term ones. We could not ask Marvell for > > something they could not give us. > > Well, your second paragraph just made it abundantly clear that Marvell > will never release documentation or a distributable firmware image. I said that they cannot distribute their current firmware, because they do not own the rights to part of it. I did not say that firmware could not be released someday, just that their current firmware is entangled in ways that they *cannot* quickly disentangle. We deal with the world the way it is, rather than the way we'd like it to be. > I > have heard this same excuse from many American wireless chip vendors > (and never from any non-American wireless chip vendors). And yet, > since you did not know the lines they would play you with, you have > not forced them to write a replacement firmware, nor have you forced > them to ever hand out documentation to get something like that > written. You have gotten nothing but a specially hacked driver, > helped by one or two people at Marvell, running with a special > firmware which Red Hat will then sign a contract to get distribution > rights for (and noone else will be able to distribute it). Am I not > correct? Do you really see it going anywhere else? > > Furthermore, it seems you are saying that the OLPC will not be based > on fully open or free software in the end. And you are saying that > your financial clout of millions of chips was not enough to put you in > ANY position of power. No, I was saying our options was limited from the start of the reality of Marvell's code base, and the lack of alternatives. There was not time to start from a clean slate for the firmware for the part. > > The OLPC project from the start gained a good reputation in the > open/free software community from acting as if it will be open & free; > the fact is it is not based on the open & free principles of our > greater community. People understood that this would all be based on > free stuff, and that source produced for this project will all find > it's way back into the ecosystem. But that is not a requirement of your > process -- it has been downgraded to something you only hope for. > > Why don't you tell people the truth -- that OLPC will very likely > depend on closed parts. That currently there are expedient steps > being taken that may not gaurantee a fully open platform in the end. > If you will not, should I go make public statements about it? I am > willing to explain the problem to the masses for you. You can believe me when I say that we have not let the issue drop with Marvell, or you may not. I repeat, we have not let the issue drop. Here's a bit more detail of what is going on behind the scenes: we do have a plan underway to start by getting an open version of the MIT Roofnet software (the mesh networking firmware) out. Michailis Bletsas here is organizing this effort. This is a start, but not the end point that either you or us desire, and we do not expect the work to end there. > > > You also need to understand that this component was/is literally the > > only game in town to achieve our goals: *there are no other > > alternatives*. We had no choice at all. Not Intel. Not Broadcom. Not > > Athleros. That Marvell chip is currently a unique part, *and they know > > it*. Negotiating positions are much stronger when you have > > alternatives, and we did not. For just about everything else, we were > > able to negotiate from strength; but not for this component. > > Frankly, now that our team has reverse engineered a multitude of > wireless chips, I am sure you are talking with any knowledge of the > domain. To pick just one example of a choice you had you could have > funded the prism54 people (who already have the skills) to build a > replacement firmware for any other chip vendor you partially opened up > with your financial clout, since about half of the other chipset > vendors also have arm or mips cpus on their 802.11 chips, which can > run even when the host is turned off. There is nothing unique about > the Marvell chip. It's just a processor that can can run when the > main cpu is turned off, and that just requires that USB or other power > be maintained. That's not a big deal. At a price point we could afford, *there are no alternatives right now*. It is in fact a unique part at anything like the price we got, working at the power consumption levels it does. If there are people interested in helping the open firmware effort, Michailis Bletsas is the person to contact. > > According to the latest estimates you are publishing, you are talking > at least 8 million Marvell chips. And you gain nothing except two > minor slaves at Marvell? > > I think you went from great potential strength to absolute weakness in > dealing with a vendor; used the open source community to gather > momentum you did not deserve; have hurt our efforts at working with > other vendors. And you did not even apologize for that damaging > effect you are having on the open/free source community's efforts to > free up more chips! > > Quite frankly the typical free/open operating system user needs good > wireless support from their operating system more than some commercial > spin-off project from Red Hat needs support for one locked-down chip. > Or have you forgotten that we were supposed to be a community? > > Yes, it sounds pretty harsh when stated that way. But can I ask > precisely when you lost your zeal to even _try_ to live up to the > commonly held goals of the community? In the end, OLPC is a project to help the children of the world to *learn*. Free and open software is a means to an end, rather than the sole end unto itself for OLPC. It is an important aspect of kids learning that they be able to see how computers work, but we had to choose getting systems into the hands of kids where almost all source is available sooner, or not getting systems out at all for months longer (later), or which would not be usable with human power that is needed for a large fraction of the kids on the planet, what did you expect our priorities to be? Regards, - Jim Gettys -- Jim Gettys One Laptop Per Child Replied: Thu, 05 Oct 2006 11:31:28 -0600 Replied: "rms@gnu.org jg@laptop.org, jsg@goblin.cx, dcbw@redhat.com, mtosatti@redhat.com" Return-Path: rms@gnu.org Delivery-Date: Thu Oct 5 10:45:59 2006 Received: from fencepost.gnu.org (fencepost.gnu.org [199.232.76.164]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k95Gjxk9029590 for ; Thu, 5 Oct 2006 10:45:59 -0600 (MDT) Received: from rms by fencepost.gnu.org with local (Exim 4.34) id 1GVWLs-0002y4-HY; Thu, 05 Oct 2006 12:45:40 -0400 Content-Type: text/plain; charset=ISO-8859-15 From: Richard Stallman To: jg@laptop.org CC: deraadt@cvs.openbsd.org, jsg@goblin.cx, dcbw@redhat.com, mtosatti@redhat.com, rms@gnu.org In-reply-to: <1159989002.6029.710.camel@localhost.localdomain> (message from Jim Gettys on Wed, 04 Oct 2006 15:10:02 -0400) Subject: Re: Marvell 88W8388 documentation Reply-to: rms@gnu.org References: <200610041838.k94IcY6j006189@cvs.openbsd.org> <1159989002.6029.710.camel@localhost.localdomain> Message-Id: Date: Thu, 05 Oct 2006 12:45:40 -0400 And there is a reality that a piece of this is out of Marvell's short term control: their firmware was based on a commercial embedded OS, I presume you mean that this embedded OS is proprietary. Whether it is commercial (developed as a business) is neither here nor there. See http://www.gnu.org/philosophy/words-to-avoid.html#Commercial. not theirs to give away. Whether they release the code is not the crucial issue. What we need above all is the full specs of their chip, not under NDA. With that, we could write our own code. Sure, it would be even better if they release the complete code of the program that runs. That would be a bonus. But we don't _need_ that. An intermediate option would be to release only their part of the code. This would be a partial bonus, nicer than no bonus at all. But still not necessary. It is well to ask them for a bonus if you think they might give it. But let's keep the focus on the crucial issue: getting what we really need. Negotiating positions are much stronger when you have alternatives, and we did not. For just about everything else, we were able to negotiate from strength; but not for this component. The alternative I would have chosen is to give up the mesh feature. I would choose that rather than use non-free software. That would be a substantial sacrifice, but sometimes freedom requires a sacrifice. The willingness to make such sacrifices would increase your negotiating power to get what you want without a sacrifice. In the longer term, there may be other potential alternatives that would be less of a sacrifice. Are there any other companies interested in developing a part that could compete with this one? Getting in on the plans early, you could work with them to get the right outcome. Return-Path: rms@gnu.org Delivery-Date: Thu Oct 5 10:46:20 2006 Received: from fencepost.gnu.org (fencepost.gnu.org [199.232.76.164]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k95GkK7x014606 for ; Thu, 5 Oct 2006 10:46:20 -0600 (MDT) Received: from rms by fencepost.gnu.org with local (Exim 4.34) id 1GVWLz-0002z5-0u; Thu, 05 Oct 2006 12:45:47 -0400 Content-Type: text/plain; charset=ISO-8859-15 From: Richard Stallman To: Theo de Raadt CC: jg@laptop.org, jsg@goblin.cx, dcbw@redhat.com, mtosatti@redhat.com cc: rms@gnu.org In-reply-to: <200610041941.k94JfaPt019572@cvs.openbsd.org> (message from Theo de Raadt on Wed, 04 Oct 2006 13:41:36 -0600) Subject: Re: Marvell 88W8388 documentation Reply-to: rms@gnu.org References: <200610041941.k94JfaPt019572@cvs.openbsd.org> Message-Id: Date: Thu, 05 Oct 2006 12:45:47 -0400 The OLPC project from the start gained a good reputation in the open/free software community from acting as if it will be open & free; the fact is it is not based on the open & free principles of our greater community. Hear, hear! Return-Path: deraadt@cvs.openbsd.org Delivery-Date: Thu Oct 5 11:49:51 2006 Received: from cvs.openbsd.org (localhost [127.0.0.1]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k95HnnNs012837 for ; Thu, 5 Oct 2006 11:49:51 -0600 (MDT) Message-Id: <200610051749.k95HnnNs012837@cvs.openbsd.org> Date: Thu, 05 Oct 2006 11:49:49 -0600 From: Theo de Raadt Subject: Letter to OLPC BCC: ------- Blind-Carbon-Copy To: lwn@lwn.net, jeremy@kerneltrap.org, sam@gnubies.com, jem@thejemreport.com Subject: Letter to OLPC Date: Thu, 05 Oct 2006 11:49:49 -0600 From: Theo de Raadt I have decided to make public this letter which I sent to the OLPC ("One Laptop Per Child" group, which is strongly associated with Red Hat. There have been replies to it by both Jim Gettys (argueing that their expediency is justified) and RMS (agreeing strongly with my point of view), but I will not disclose their letters. I am getting really tired of "open source" people who work against the open source community. Our little group can probably take credit for having "opened up" more wireless devices than the rest of the community, and therefore we feel we have a better grasp of the damage OLPC has done here. Our reverse engineering and documentation efforts will in time help all free software projects. Please take note, and publish if you wish. Thanks. - --- To: Jonathan Gray cc: dcbw@redhat.com, jg@laptop.org, mtosatti@redhat.com cc: rms@gnu.org cc: deraadt Subject: Re: Marvell 88W8388 documentation In-reply-to: Your message of "Thu, 14 Sep 2006 22:47:00 +1000." <20060914124700.GA21474@mail.netspace.net.au> Date: Wed, 04 Oct 2006 12:38:34 -0600 From: Theo de Raadt > Please correct me if I am wrong but it seems that documentation > for Marvell's 88W8388's is not publically available without > signing an NDA? > > If this is the case why did a project that seems to pride > itself of openess agree to deal with such a company? > Drivers written under NDA tend to be full of magic numbers, > near impossible for others to properly maintain and > totally against the spirit of open projects. > > I really think you should push for Marvell to give out > documentation without them forcing NDAs onto people. > Failing that I'm sure there are other vendors > who would be willing to be more helpful. Jonathan showed me this mail he sent you about your NDA "cooperation" with Marvell for the wireless chip that you want to use for the OLPC project, so that Marvell will write you special hacks to do low-power mesh networking while the main cpu is powered off. This does not gaurantee Marvell is going to be open and release documentation for their chips though. When large players like you make such private agreements with such secretive vendors, you work against our common goals of getting more open documentation for devices. It is only with open documentation that OS groups can increase device support, and later -- keep the device drivers reliable after the device is EOL'd by the vendor. I've heard claims that you (OLPC members, Red Hat employees) think this relationship with Marvell will eventually prompt/teach them to be more open in time. Do you not realize how much of a DELUSION the history of free/open operating systems shows that point of view to be? Very few chip vendors have ever opened up unless they were pushed, let alone Marvell (who I am led to believe also has NDA's with Red Hat employees for the Marvell Yukon/Yukon 2 gigabit ethernet chips -- again one of the few closed chips). It is clear that your choices are not about opening up Marvell, but simply commercially expedient and hurtful to our common cause. You came to Marvell with potential sales of millions of units, and then completely wimped out in demanding ideals that you say you share with the community. Now other companies like Intel, Broadcom, and TI can say to us "Why should we open up, Marvell did not have to". So I must say I am extremely dissapointed you have chosen to work against the very obvious goals of "open", and I hope that in time you are made to feel ashamed of the choice you have made. ------- End of Blind-Carbon-Copy Return-Path: owner-misc+M37805@openbsd.org Delivery-Date: Thu Oct 5 11:58:32 2006 Received: from shear.ucar.edu (shear.ucar.edu [192.43.244.163]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k95HwV74009563; Thu, 5 Oct 2006 11:58:32 -0600 (MDT) Received: from openbsd.org (localhost.ucar.edu [127.0.0.1]) by shear.ucar.edu (8.13.8/8.13.6) with ESMTP id k95HsCIt004561; Thu, 5 Oct 2006 11:54:12 -0600 (MDT) Received: from cvs.openbsd.org (cvs.openbsd.org [199.185.137.3]) by shear.ucar.edu (8.13.8/8.13.6) with ESMTP id k95HotIb011263 (version=TLSv1/SSLv3 cipher=DHE-DSS-AES256-SHA bits=256 verify=FAIL) for ; Thu, 5 Oct 2006 11:50:55 -0600 (MDT) Received: from cvs.openbsd.org (localhost [127.0.0.1]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k95Hos0I019862 for ; Thu, 5 Oct 2006 11:50:54 -0600 (MDT) Message-Id: <200610051750.k95Hos0I019862@cvs.openbsd.org> To: misc@cvs.openbsd.org Subject: Letter to OLPC Date: Thu, 05 Oct 2006 11:50:54 -0600 From: Theo de Raadt X-Loop: misc@openbsd.org Precedence: list Sender: owner-misc@openbsd.org I have decided to make public this letter which I sent to the OLPC ("One Laptop Per Child" group, which is strongly associated with Red Hat. There have been replies to it by both Jim Gettys (argueing that their expediency is justified) and RMS (agreeing strongly with my point of view), but I will not disclose their letters. I am getting really tired of "open source" people who work against the open source community. Our little group can probably take credit for having "opened up" more wireless devices than the rest of the community, and therefore we feel we have a better grasp of the damage OLPC has done here. Our reverse engineering and documentation efforts will in time help all free software projects. Please take note, and publish if you wish. Thanks. --- To: Jonathan Gray cc: dcbw@redhat.com, jg@laptop.org, mtosatti@redhat.com cc: rms@gnu.org cc: deraadt Subject: Re: Marvell 88W8388 documentation In-reply-to: Your message of "Thu, 14 Sep 2006 22:47:00 +1000." <20060914124700.GA21474@mail.netspace.net.au> Date: Wed, 04 Oct 2006 12:38:34 -0600 From: Theo de Raadt > Please correct me if I am wrong but it seems that documentation > for Marvell's 88W8388's is not publically available without > signing an NDA? > > If this is the case why did a project that seems to pride > itself of openess agree to deal with such a company? > Drivers written under NDA tend to be full of magic numbers, > near impossible for others to properly maintain and > totally against the spirit of open projects. > > I really think you should push for Marvell to give out > documentation without them forcing NDAs onto people. > Failing that I'm sure there are other vendors > who would be willing to be more helpful. Jonathan showed me this mail he sent you about your NDA "cooperation" with Marvell for the wireless chip that you want to use for the OLPC project, so that Marvell will write you special hacks to do low-power mesh networking while the main cpu is powered off. This does not gaurantee Marvell is going to be open and release documentation for their chips though. When large players like you make such private agreements with such secretive vendors, you work against our common goals of getting more open documentation for devices. It is only with open documentation that OS groups can increase device support, and later -- keep the device drivers reliable after the device is EOL'd by the vendor. I've heard claims that you (OLPC members, Red Hat employees) think this relationship with Marvell will eventually prompt/teach them to be more open in time. Do you not realize how much of a DELUSION the history of free/open operating systems shows that point of view to be? Very few chip vendors have ever opened up unless they were pushed, let alone Marvell (who I am led to believe also has NDA's with Red Hat employees for the Marvell Yukon/Yukon 2 gigabit ethernet chips -- again one of the few closed chips). It is clear that your choices are not about opening up Marvell, but simply commercially expedient and hurtful to our common cause. You came to Marvell with potential sales of millions of units, and then completely wimped out in demanding ideals that you say you share with the community. Now other companies like Intel, Broadcom, and TI can say to us "Why should we open up, Marvell did not have to". So I must say I am extremely dissapointed you have chosen to work against the very obvious goals of "open", and I hope that in time you are made to feel ashamed of the choice you have made. Replied: Thu, 05 Oct 2006 14:15:01 -0600 Replied: "Dan Williams Jonathan Gray , jg@laptop.org, mtosatti@redhat.com, Replied: rms@gnu.org, Chris Blizzard " Return-Path: dcbw@redhat.com Delivery-Date: Thu Oct 5 13:04:39 2006 Received: from mx1.redhat.com (mx1.redhat.com [66.187.233.31]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k95J4cme025826 for ; Thu, 5 Oct 2006 13:04:39 -0600 (MDT) Received: from int-mx1.corp.redhat.com (int-mx1.corp.redhat.com [172.16.52.254]) by mx1.redhat.com (8.12.11.20060308/8.12.11) with ESMTP id k95J4WCG022327; Thu, 5 Oct 2006 15:04:32 -0400 Received: from mail.boston.redhat.com (mail.boston.redhat.com [172.16.76.12]) by int-mx1.corp.redhat.com (8.12.11.20060308/8.12.11) with ESMTP id k95J4Vii003713; Thu, 5 Oct 2006 15:04:31 -0400 Received: from [10.13.248.31] (vpn-248-31.boston.redhat.com [10.13.248.31]) by mail.boston.redhat.com (8.12.8/8.12.8) with ESMTP id k95J4UX2004144; Thu, 5 Oct 2006 15:04:30 -0400 Subject: Re: Marvell 88W8388 documentation From: Dan Williams To: Theo de Raadt Cc: Jonathan Gray , jg@laptop.org, mtosatti@redhat.com, rms@gnu.org, Chris Blizzard In-Reply-To: <200610041838.k94IcY6j006189@cvs.openbsd.org> References: <200610041838.k94IcY6j006189@cvs.openbsd.org> Content-Type: text/plain Date: Thu, 05 Oct 2006 15:06:34 -0400 Message-Id: <1160075195.8246.12.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Evolution 2.8.0 (2.8.0-7.fc6) Content-Transfer-Encoding: 7bit On Wed, 2006-10-04 at 12:38 -0600, Theo de Raadt wrote: > > Please correct me if I am wrong but it seems that documentation > > for Marvell's 88W8388's is not publically available without > > signing an NDA? > > > > If this is the case why did a project that seems to pride > > itself of openess agree to deal with such a company? > > Drivers written under NDA tend to be full of magic numbers, > > near impossible for others to properly maintain and > > totally against the spirit of open projects. > > > > I really think you should push for Marvell to give out > > documentation without them forcing NDAs onto people. > > Failing that I'm sure there are other vendors > > who would be willing to be more helpful. > > Jonathan showed me this mail he sent you about your NDA "cooperation" > with Marvell for the wireless chip that you want to use for the OLPC > project, so that Marvell will write you special hacks to do low-power > mesh networking while the main cpu is powered off. This does not > gaurantee Marvell is going to be open and release documentation for > their chips though. > > When large players like you make such private agreements with such > secretive vendors, you work against our common goals of getting more > open documentation for devices. It is only with open documentation > that OS groups can increase device support, and later -- keep the > device drivers reliable after the device is EOL'd by the vendor. > > I've heard claims that you (OLPC members, Red Hat employees) think > this relationship with Marvell will eventually prompt/teach them to be > more open in time. Do you not realize how much of a DELUSION the > history of free/open operating systems shows that point of view to be? > Very few chip vendors have ever opened up unless they were pushed, let > alone Marvell (who I am led to believe also has NDA's with Red Hat > employees for the Marvell Yukon/Yukon 2 gigabit ethernet chips -- > again one of the few closed chips). I would urge you to wait. You seem to expect this process to happen over night, which is clearly impossible. As I'm sure you have experienced from the OpenBSD firmware re-licensing effort, this stuff often moves at a snails pace. It is unreasonable to expect an "open & release everything now!!" request to be honored at light-speed. The end goals have not changed; freely redistributable firmware, open drivers, and documentation. Perhaps these goals were not adequately expressed to you and others, and if not, then maybe OLPC needs to try harder to express them. Dan > It is clear that your choices are not about opening up Marvell, but > simply commercially expedient and hurtful to our common cause. You > came to Marvell with potential sales of millions of units, and then > completely wimped out in demanding ideals that you say you share with > the community. Now other companies like Intel, Broadcom, and TI can > say to us "Why should we open up, Marvell did not have to". > > So I must say I am extremely dissapointed you have chosen to work > against the very obvious goals of "open", and I hope that in time you > are made to feel ashamed of the choice you have made. Return-Path: deraadt@cvs.openbsd.org Delivery-Date: Thu Oct 5 14:15:02 2006 Received: from cvs.openbsd.org (localhost [127.0.0.1]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k95KF1HA013595; Thu, 5 Oct 2006 14:15:02 -0600 (MDT) Message-Id: <200610052015.k95KF1HA013595@cvs.openbsd.org> Date: Thu, 05 Oct 2006 14:15:01 -0600 From: Theo de Raadt Subject: Re: Marvell 88W8388 documentation BCC: ------- Blind-Carbon-Copy To: Dan Williams cc: Jonathan Gray , jg@laptop.org, mtosatti@redhat.com, rms@gnu.org, Chris Blizzard Subject: Re: Marvell 88W8388 documentation In-reply-to: Your message of "Thu, 05 Oct 2006 15:06:34 EDT." <1160075195.8246.12.camel@localhost.localdomain> Date: Thu, 05 Oct 2006 14:15:01 -0600 From: Theo de Raadt > I would urge you to wait. You seem to expect this process to happen > over night, which is clearly impossible. Well, I see no point in waiting -- I think it will be harder to fix this problem later. You have already signed an agreement with Marvell, so now I am sure that full documentation will never be made available through your effort. I don't know of one case in history where signing an NDA first has resulted in free access later. I've been dealing with hardware documentation access issues with hardware vendors for NEARLY TWENTY YEARS (Yes, like back when Van Jacobson was having difficulty getting Intel 10mbit ethernet docs, in 1987). History is on our side. History is not on your side. > As I'm sure you have > experienced from the OpenBSD firmware re-licensing effort, this stuff > often moves at a snails pace. Actually, every vendor who has accomodated us did so pretty damn fast, and every vendor who has refused is still busy refusing. And why are they refusing? Because for all our community pressure on them, noone with any money has convinced them otherwise. Then someone with money shows up.... and immediately makes a secret agreement for secret access and vendor-only distribution. > It is unreasonable to expect an "open & > release everything now!!" request to be honored at light-speed. You've been talking with Marvell for long enough to be able to know whether or not full documentation plus firmware distribution rights will arrive. Jim's mail yesterday made it pretty clear that since Marvell pparently does not own their own product outright, that is super unlikely. > The end goals have not changed; freely redistributable firmware, open > drivers, and documentation. Perhaps these goals were not adequately > expressed to you and others, and if not, then maybe OLPC needs to try > harder to express them. Jim also made it totally clear yesterday that this is more about "laptops for poor kids" than about totally free & open software. Here, we are calling it "One Lock Per Child", or "One License Per Child". I've taken my concerns to the public. I do not see why you should get any street-cred by being associated with open/free software community, when you are morally bankrupt regarding licensing. You are playing the same "Intellectual Property" game as Marvell by even having ANY contract or agreement with them. I have one question. If in 2 years Marvell still won't make all the documentation available for everyone, and they still won't make the firmware 100% distributable... what will you do? Cancel your agreement and use something else? Declare negotiations with Marvell over, and say "Theo was right"? No. You won't. You will still use their non-free crap, and give up on your free-software goals. And that is why your calls for me to wait, to be patient, are the most retarded and morally bankrupt bullshit I've heard all day. ------- End of Blind-Carbon-Copy Return-Path: mtosatti@redhat.com Delivery-Date: Thu Oct 5 15:58:30 2006 Received: from hera.kernel.org (hera.kernel.org [140.211.167.34]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k95LwSDF011213 (version=TLSv1/SSLv3 cipher=DHE-DSS-AES256-SHA bits=256 verify=NO) for ; Thu, 5 Oct 2006 15:58:29 -0600 (MDT) Received: from dmt.cnet (IDENT:U2FsdGVkX1+GDYkLpuK3E+Tr6uTCbxP9V+wEaeUw6HU@localhost [127.0.0.1]) by hera.kernel.org (8.13.7/8.13.7) with ESMTP id k95JgqEp019493; Thu, 5 Oct 2006 19:42:55 GMT Received: from dmt.cnet (dmt.cnet [127.0.0.1]) by dmt.cnet (Postfix) with ESMTP id 98C004971D1; Thu, 5 Oct 2006 16:42:46 -0300 (BRT) Received: (from marcelo@localhost) by dmt.cnet (8.13.6/8.13.6/Submit) id k95JgTg6009902; Thu, 5 Oct 2006 16:42:29 -0300 Date: Thu, 5 Oct 2006 16:42:28 -0300 From: Marcelo Tosatti To: Richard Stallman Cc: jg@laptop.org, deraadt@cvs.openbsd.org, jsg@goblin.cx, dcbw@redhat.com Subject: Re: Marvell 88W8388 documentation Message-ID: <20061005194228.GA9787@dmt> References: <200610041838.k94IcY6j006189@cvs.openbsd.org> <1159989002.6029.710.camel@localhost.localdomain> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.4.2.1i X-Spam-Status: No, score=-3.7 required=5.0 tests=ALL_TRUSTED,AWL,BAYES_00 autolearn=ham version=3.1.3 X-Spam-Checker-Version: SpamAssassin 3.1.3 (2006-06-01) on hera.kernel.org On Thu, Oct 05, 2006 at 12:45:40PM -0400, Richard Stallman wrote: > And there is a reality that a piece of this is out of Marvell's short > term control: their firmware was based on a commercial embedded OS, > > I presume you mean that this embedded OS is proprietary. Whether it > is commercial (developed as a business) is neither here nor there. > See http://www.gnu.org/philosophy/words-to-avoid.html#Commercial. > > not > theirs to give away. > > Whether they release the code is not the crucial issue. What we need > above all is the full specs of their chip, not under NDA. With that, > we could write our own code. Certainly thats what needs to be done as a practical step forward. As it stands now, Christopher/Michail/Jim have the power to articulate open documentation requirements. Has the OLPC project ever requested such thing, and what was the outcome? > Sure, it would be even better if they release the complete code of the > program that runs. That would be a bonus. But we don't _need_ that. > > An intermediate option would be to release only their part of the > code. This would be a partial bonus, nicer than no bonus at all. > But still not necessary. > > It is well to ask them for a bonus if you think they might give it. > But let's keep the focus on the crucial issue: getting what we really > need. > > Negotiating positions are much stronger when you have > alternatives, and we did not. For just about everything else, we were > able to negotiate from strength; but not for this component. > > The alternative I would have chosen is to give up the mesh feature. I > would choose that rather than use non-free software. That would be a > substantial sacrifice, but sometimes freedom requires a sacrifice. > > The willingness to make such sacrifices would increase your > negotiating power to get what you want without a sacrifice. > > In the longer term, there may be other potential alternatives that > would be less of a sacrifice. Are there any other companies > interested in developing a part that could compete with this one? > Getting in on the plans early, you could work with them to get the > right outcome. Return-Path: deraadt@cvs.openbsd.org Delivery-Date: Thu Oct 5 16:16:32 2006 Received: from cvs.openbsd.org (localhost [127.0.0.1]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k95MGWWM031339; Thu, 5 Oct 2006 16:16:32 -0600 (MDT) Message-Id: <200610052216.k95MGWWM031339@cvs.openbsd.org> Date: Thu, 05 Oct 2006 16:16:32 -0600 From: Theo de Raadt Subject: Re: Marvell 88W8388 documentation BCC: ------- Blind-Carbon-Copy To: jg@laptop.org cc: Jonathan Gray , dcbw@redhat.com, mtosatti@redhat.com, rms@gnu.org Subject: Re: Marvell 88W8388 documentation In-reply-to: Your message of "Wed, 04 Oct 2006 16:20:36 EDT." <1159993236.6029.758.camel@localhost.localdomain> Date: Thu, 05 Oct 2006 16:16:32 -0600 From: Theo de Raadt I wanted to show you all a mail that one very astute person on the misc@openbsd.org mailing list posted. He details precisely how OLPC / Red Hat should have approached this matter. But that is not how it was done. Instead, you behaved towards Marvell from a position of supreme supplication, doing nothing more than begging to lick at their teat. If Red Hat -- the strongman of the open source community -- continues to act like absolute wimps, how do you think that makes the little players like us look? Please grab some balls and go back to Marvell and renegotiate this deal immediately -- for *FULL* documentation disclosure and firmware distribution licensing as soon as possible. - -- Date: Fri, 6 Oct 2006 00:08:24 +0200 From: Paul de Weerd To: "Jack J. Woehr" Cc: OpenBSD Subject: Re: Letter to OLPC On Thu, Oct 05, 2006 at 03:54:47PM -0600, Jack J. Woehr wrote: | > Free and open software is a means to an end, rather than the | > sole end unto itself for OLPC. | > | > I was totally stunned by this admission. "morally bankrupt", as Bob | > says, is exactly what is going on. | | Hmm, sounds like you are saying that abstract goal of unlimited | software freedom is | a higher goal than providing access to modern technology to | disadvantaged children in | 3rd-world countries. I don't wish to argue that point, but it is | certainly a point | that could be debated. Why *would* the OLPC people wish to get their | dicks caught | in the struggle between the free-and-open software community and the | greedheads? This is a perfect opportunity to stand up, speak up about this issue. Why would the Intels and Marvells of this world withhold developers the documentation they need if they are unwilling to sign an NDA ? They are writing software that provides 'disadvantaged children in 3rd-world countries' access to modern technology. Reverse your argument and bring it to Marvell. Imagine the bad press Marvell would have gotten had they declined OLPC/Red Hat access to the documentation without NDA when asked. "This company will not allow 'disadvantaged children in 3rd-world countries' to gain access to modern technology, because they feel the documentation to their hardware is to secret." (or whatever their false reasoning is) What these companies need is bad press. Bad press is bad for their business and shareholders will start to complain. It seems that this is the only way to make changes in big corporations, and changes are exactly what we need. Paul 'WEiRD' de Weerd ------- End of Blind-Carbon-Copy Replied: Thu, 05 Oct 2006 17:33:40 -0600 Replied: "jg@laptop.org rms@gnu.org, jsg@goblin.cx, dcbw@redhat.com, mtosatti@redhat.com" Return-Path: jg@laptop.org Delivery-Date: Thu Oct 5 17:00:13 2006 Received: from alnrmhc11.comcast.net (alnrmhc11.comcast.net [206.18.177.51]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k95N0DuG020457 for ; Thu, 5 Oct 2006 17:00:13 -0600 (MDT) Received: from wireless-19-154.media.mit.edu ([18.85.19.154]) by comcast.net (alnrmhc11) with SMTP id <20061005224455b11009kpste>; Thu, 5 Oct 2006 22:44:56 +0000 Subject: Re: Marvell 88W8388 documentation From: Jim Gettys Reply-To: jg@laptop.org To: rms@gnu.org Cc: deraadt@cvs.openbsd.org, jsg@goblin.cx, dcbw@redhat.com, mtosatti@redhat.com In-Reply-To: References: <200610041838.k94IcY6j006189@cvs.openbsd.org> <1159989002.6029.710.camel@localhost.localdomain> Content-Type: text/plain Organization: OLPC Date: Thu, 05 Oct 2006 18:44:53 -0400 Message-Id: <1160088294.6075.40.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Evolution 2.6.1 Content-Transfer-Encoding: 7bit On Thu, 2006-10-05 at 12:45 -0400, Richard Stallman wrote: > And there is a reality that a piece of this is out of Marvell's short > term control: their firmware was based on a commercial embedded OS, > > I presume you mean that this embedded OS is proprietary. Whether it > is commercial (developed as a business) is neither here nor there. > See http://www.gnu.org/philosophy/words-to-avoid.html#Commercial. > > not > theirs to give away. > > Whether they release the code is not the crucial issue. What we need > above all is the full specs of their chip, not under NDA. With that, > we could write our own code. > > Sure, it would be even better if they release the complete code of the > program that runs. That would be a bonus. But we don't _need_ that. > > An intermediate option would be to release only their part of the > code. This would be a partial bonus, nicer than no bonus at all. > But still not necessary. > > It is well to ask them for a bonus if you think they might give it. > But let's keep the focus on the crucial issue: getting what we really > need. > > Negotiating positions are much stronger when you have > alternatives, and we did not. For just about everything else, we were > able to negotiate from strength; but not for this component. > > The alternative I would have chosen is to give up the mesh feature. I > would choose that rather than use non-free software. That would be a > substantial sacrifice, but sometimes freedom requires a sacrifice. > > The willingness to make such sacrifices would increase your > negotiating power to get what you want without a sacrifice. There is no power for access points in the parts of the world we go to. This is not a viable option. Most of the kids in the developing world have no electricity, much less money for access points. The only way we have to solve the last mile problem at all is via a mesh network. And in terms of mesh networking, the alternative was Atheros, with their closed HAL. It is a non-starter as it cannot run without the processor running, which would require us to run at about 5-10 times the power. > > In the longer term, there may be other potential alternatives that > would be less of a sacrifice. Are there any other companies > interested in developing a part that could compete with this one? > Getting in on the plans early, you could work with them to get the > right outcome. > Yes, what do you think we are doing now for our next generation? We talk to all the manufacturers. We *want* them to compete for the business. And yes, we expect the next time around we'll have more than one possible choice. - Jim -- Jim Gettys One Laptop Per Child Replied: Thu, 05 Oct 2006 17:58:17 -0600 Replied: "jg@laptop.org Dan Williams , Jonathan Gray , Replied: mtosatti@redhat.com, rms@gnu.org, Replied: Chris Blizzard " Replied: Thu, 05 Oct 2006 17:40:40 -0600 Replied: "jg@laptop.org Dan Williams , Jonathan Gray , Replied: mtosatti@redhat.com, rms@gnu.org, Replied: Chris Blizzard " Return-Path: jg@laptop.org Delivery-Date: Thu Oct 5 17:06:57 2006 Received: from alnrmhc14.comcast.net (alnrmhc14.comcast.net [204.127.225.94]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k95N6vN5027959 for ; Thu, 5 Oct 2006 17:06:57 -0600 (MDT) Received: from wireless-19-154.media.mit.edu ([18.85.19.154]) by comcast.net (alnrmhc14) with SMTP id <20061005230640b1400kgos6e>; Thu, 5 Oct 2006 23:06:51 +0000 Subject: Re: Marvell 88W8388 documentation From: Jim Gettys Reply-To: jg@laptop.org To: Theo de Raadt Cc: Dan Williams , Jonathan Gray , mtosatti@redhat.com, rms@gnu.org, Chris Blizzard In-Reply-To: <200610052015.k95KF1H8013595@cvs.openbsd.org> References: <200610052015.k95KF1H8013595@cvs.openbsd.org> Content-Type: text/plain Organization: OLPC Date: Thu, 05 Oct 2006 19:06:38 -0400 Message-Id: <1160089598.6075.63.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Evolution 2.6.1 Content-Transfer-Encoding: 7bit On Thu, 2006-10-05 at 14:15 -0600, Theo de Raadt wrote: > > I would urge you to wait. You seem to expect this process to happen > > over night, which is clearly impossible. > > Well, I see no point in waiting -- I think it will be harder to fix > this problem later. You have already signed an agreement with > Marvell, so now I am sure that full documentation will never be made > available through your effort. This does not follow. It is an assertion without proof. > I don't know of one case in history > where signing an NDA first has resulted in free access later. Then you don't know history. Let's see: some of my personal history, which you will probably dismiss: As an employee of Digital Equipment Corporation, having signed the Digital employee agreement, I was personally responsible for causing Digital to release the programming specifications for the MIPS based workstations and Turbochannel cards. In my long experience on the other side of the fence, often the documentation doesn't really exist, are in some random directory where no-one knows where it is, or is so intertwined with hardware documentation that extricating programming specifications from those documents is a long and arduous process. Very often it is out-dated and incorrect. Sometimes, it is entwined with future product plans and specifications. The reason that the MIPS documents happened is that I bothered to track them down and clean them up enough to get them out there, and also that the engineering group involved was better than average at doing documentation. I spent more than a month getting those documents in shape that the could be released, personally. > I've > been dealing with hardware documentation access issues with hardware > vendors for NEARLY TWENTY YEARS (Yes, like back when Van Jacobson was > having difficulty getting Intel 10mbit ethernet docs, in 1987). > History is on our side. History is not on your side. See above history. My history is longer than yours: I've been at this since 1983; so I'm at 23 years and counting. I worked on drafting the MIT license. Please don't tell me I don't know anything about this topic. > > > As I'm sure you have > > experienced from the OpenBSD firmware re-licensing effort, this stuff > > often moves at a snails pace. > > Actually, every vendor who has accomodated us did so pretty damn fast, > and every vendor who has refused is still busy refusing. And why are > they refusing? Because for all our community pressure on them, noone > with any money has convinced them otherwise. Actually, from what I've seen, community pressure has had at best mixed success at best. What actually works is economics and self interest, and pointing out the economics and self interest to those companies. > Then someone with money > shows up.... and immediately makes a secret agreement for secret > access and vendor-only distribution. You can often get more by being pleasant than confrontational. I don't know what you mean by vendor-only distribution. > > > It is unreasonable to expect an "open & > > release everything now!!" request to be honored at light-speed. > > You've been talking with Marvell for long enough to be able to know > whether or not full documentation plus firmware distribution rights > will arrive. Jim's mail yesterday made it pretty clear that since > Marvell pparently does not own their own product outright, that is > super unlikely. They do not own the simple operating system they used for their firmware. As I said in my last mail, we already have work underway to be to replace this firmware. You can believe this statement, or not, but it is true. You can help, but apparently you'd prefer to not follow up and just write more mail messages. > > > The end goals have not changed; freely redistributable firmware, open > > drivers, and documentation. Perhaps these goals were not adequately > > expressed to you and others, and if not, then maybe OLPC needs to try > > harder to express them. > > Jim also made it totally clear yesterday that this is more about > "laptops for poor kids" than about totally free & open software. > Here, we are calling it "One Lock Per Child", or "One License Per > Child". Yes, I DO NOT APOLOGIZE FOR PUTTING THE GOOD OF THE WORLDS CHILDREN FIRST. We are One Laptop Per Child.... > > I've taken my concerns to the public. I do not see why you should get > any street-cred by being associated with open/free software community, > when you are morally bankrupt regarding licensing. You are playing > the same "Intellectual Property" game as Marvell by even having ANY > contract or agreement with them. > > I have one question. If in 2 years Marvell still won't make all the > documentation available for everyone, and they still won't make the > firmware 100% distributable... what will you do? Cancel your agreement > and use something else? Two years from now we'll be using different parts and have more options. No machine stays in production for 2 years. > Declare negotiations with Marvell over, and > say "Theo was right"? > > No. You won't. You will still use their non-free crap, and give up > on your free-software goals. And that is why your calls for me to > wait, to be patient, are the most retarded and morally bankrupt > bullshit I've heard all day. Our goals are helping kids learn. If it is so terrible to have built the most open system in recent memory, while continuing to do work freeing up the firmware on the one part that the one available component that can do what we need is underway, then I guess by your definition, we must be terrible. Ad homonym, abusive attacks on people's motives, goals or future behavior do not help your case one bit. -- Jim Gettys One Laptop Per Child Return-Path: deraadt@cvs.openbsd.org Delivery-Date: Thu Oct 5 18:31:16 2006 Received: from cvs.openbsd.org (localhost [IPv6:::1]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k960VGQk005674 (version=TLSv1/SSLv3 cipher=DHE-DSS-AES256-SHA bits=256 verify=NO) for ; Thu, 5 Oct 2006 18:31:16 -0600 (MDT) Received: (from deraadt@localhost) by cvs.openbsd.org (8.13.6/8.12.0/Submit) id k960VGoI017737 for deraadt; Thu, 5 Oct 2006 18:31:16 -0600 (MDT) Date: Thu, 5 Oct 2006 18:31:16 -0600 (MDT) From: Theo de Raadt Message-Id: <200610060031.k960VGoI017737@cvs.openbsd.org> To: deraadt@cvs.openbsd.org To: jg@laptop.org cc: rms@gnu.org, jsg@goblin.cx, dcbw@redhat.com, mtosatti@redhat.com Subject: Re: Marvell 88W8388 documentation In-reply-to: Your message of "Thu, 05 Oct 2006 18:44:53 EDT." <1160088294.6075.40.camel@localhost.localdomain> Date: Thu, 05 Oct 2006 17:33:40 -0600 From: Theo de Raadt > There is no power for access points in the parts of the world we go to. > This is not a viable option. > > Most of the kids in the developing world have no electricity, much less > money for access points. The only way we have to solve the last mile > problem at all is via a mesh network. [The above text is completely unrelated to how you negotiated weakly with Marvell. You keep bringing it up as the excuse for why the Marvell agreement was made, but it simply does not matter in the big picture.] > And in terms of mesh networking, the alternative was Atheros, with their > closed HAL. It is a non-starter as it cannot run without the processor > running, which would require us to run at about 5-10 times the power. For the record, we have completely reverse engineered the Atheros HAL ourselves. We've been using our own completely open code for four years. So there was an option. Maybe it used more power. So? I bet you were not even aware of our code. Of course, you would never have had the balls to ask Atheros to open up their HAL, either. And by the way you can find 2-year old offical Atheros HAL source code on some sites in eastern Europe -- with a BSD license on every file, placed there by Sam Leffler himself. It is already free code according to the license, but still, we avoided it and wrote our own. > > In the longer term, there may be other potential alternatives that > > would be less of a sacrifice. Are there any other companies > > interested in developing a part that could compete with this one? > > Getting in on the plans early, you could work with them to get the > > right outcome. > > > > Yes, what do you think we are doing now for our next generation? > > We talk to all the manufacturers. We *want* them to compete for the > business. And yes, we expect the next time around we'll have more than > one possible choice. Your mail does a very good job of convincing me that for now you MUST use the Marvell chip. Technologically, you simply do not have any other choice. The quantity of sales you may bring made no difference. And as I can see, you are already investing engineering effort into using the Marvell chips. So in the end, even if Marvell refuses to open up their documentation, you will accept their closed terms, and buy 8,000,000 of their chips, and then ship closed laptops to kids. And quite frankly, you just convinced me that Marvell will never give us any documentation either, since I know of some other large vendors who Marvell refused to give documentation to (but we also know some who Marvell did give the same documentation to). Even medium sized vendors have learned that they must DEMAND that documentation arrives before starting any engineering in-house that financially commits them towards buying a particular product. So why not simply tell Marvell that you are completely captive to their closed model? Why not honestly tell the open source community that too? Why keep trying to dance around the issues I brought up? Therefore, in response to your solid arguments that you MUST use the Marvell chip... I have a request. Can I have permission to forward your mail to Marvell? I want to show them how convincingly you argue that you are a captive customer of theirs. Maybe they will even increase the price of their chips. Now you will think I'm a bigger asshole than before. But how did you get to the point where I could even ask such a thing? Or should I just show your mail to Marvell anyways, without permission? If you are willing to scuttle our efforts to free documentation, why should we not scuttle you? Come on! Do you really not get it? If there is one lesson we should have by now it is that YOU CAN'T HAVE IT BOTH WAYS! It's the singular problem of "vendor lock-in" that caused most of us to get into this whole culture of writing free software! Is Red Hat trying to become as closed as Sun? It took them years to dig through all their paperwork to get Solaris source out the door. You want that? You would sacrifice freedom hoping to gain it back at a later date. Has that retarded American ethos really now penetrated all the way into the core of the Linux community? The last thing we expected was a bunch of woosy traitors in our midst! What DO you stand for? You cannot honestly stand only for "laptops for children" and throw away everything else that our community depends on? Should we be asking how else are you undermining the free/open source community behind our backs? I -- and many others -- are going to hold you accountable to fix this bloody mess you have created! Replied: Fri, 06 Oct 2006 14:42:07 -0600 Replied: "Marcelo Tosatti jg@laptop.org, Dan Williams , Replied: Jonathan Gray , rms@gnu.org, Replied: Chris Blizzard " Replied: Fri, 06 Oct 2006 14:18:19 -0600 Replied: "Marcelo Tosatti jg@laptop.org, Dan Williams , Replied: Jonathan Gray , rms@gnu.org, Replied: Chris Blizzard " Replied: Fri, 06 Oct 2006 14:09:16 -0600 Replied: "Marcelo Tosatti jg@laptop.org, Dan Williams , Replied: Jonathan Gray , rms@gnu.org, Replied: Chris Blizzard " Return-Path: mtosatti@redhat.com Delivery-Date: Fri Oct 6 12:35:08 2006 Received: from hera.kernel.org (hera.kernel.org [140.211.167.34]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k96IZ7pX016263 (version=TLSv1/SSLv3 cipher=DHE-DSS-AES256-SHA bits=256 verify=NO) for ; Fri, 6 Oct 2006 12:35:08 -0600 (MDT) Received: from dmt.cnet (IDENT:U2FsdGVkX1/LE11ZfjykMkylsPbC1vCE2a9ihvSyiMU@localhost [127.0.0.1]) by hera.kernel.org (8.13.7/8.13.7) with ESMTP id k96IYshS003142; Fri, 6 Oct 2006 18:34:56 GMT Received: from dmt.cnet (dmt.cnet [127.0.0.1]) by dmt.cnet (Postfix) with ESMTP id 3BF704971D1; Fri, 6 Oct 2006 14:34:20 -0300 (BRT) Received: (from marcelo@localhost) by dmt.cnet (8.13.6/8.13.6/Submit) id k96HY3Tf004865; Fri, 6 Oct 2006 14:34:03 -0300 Date: Fri, 6 Oct 2006 14:34:03 -0300 From: Marcelo Tosatti To: Theo de Raadt Cc: jg@laptop.org, Dan Williams , Jonathan Gray , rms@gnu.org, Chris Blizzard Subject: Re: Marvell 88W8388 documentation Message-ID: <20061006173402.GA4575@dmt> References: <1160089598.6075.63.camel@localhost.localdomain> <200610052358.k95NwHB2006381@cvs.openbsd.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <200610052358.k95NwHB2006381@cvs.openbsd.org> User-Agent: Mutt/1.4.2.1i X-Spam-Status: No, score=-3.9 required=5.0 tests=ALL_TRUSTED,AWL,BAYES_00 autolearn=ham version=3.1.3 X-Spam-Checker-Version: SpamAssassin 3.1.3 (2006-06-01) on hera.kernel.org On Thu, Oct 05, 2006 at 05:58:17PM -0600, Theo de Raadt wrote: > > In my long experience on the other side of the fence, often the > > documentation doesn't really exist, are in some random directory where > > no-one knows where it is, or is so intertwined with hardware > > documentation that extricating programming specifications from those > > documents is a long and arduous process. Very often it is out-dated and > > incorrect. Sometimes, it is entwined with future product plans and > > specifications. > > The Marvell documentation does exist. I have been told a bit about > how large it is. It sounds like you are making these statements > without having access to it. Its rather small (its the interface to the firmware through a set of well defined commands). By the way, you can easily understand the firmware interface by looking at the Linux driver (it provides as much information as the documentation). Everybody in the OLPC project, including the people who decided to pick a piece of hardware able to keep up with cost/power constraints, have the best intentions with reference to using _only_ free software in the machines. Replied: Fri, 06 Oct 2006 14:15:47 -0600 Replied: "rms@gnu.org jg@laptop.org, jsg@goblin.cx, dcbw@redhat.com, mtosatti@redhat.com" Return-Path: rms@gnu.org Delivery-Date: Fri Oct 6 14:06:54 2006 Received: from fencepost.gnu.org (fencepost.gnu.org [199.232.76.164]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k96K6rdo028285 for ; Fri, 6 Oct 2006 14:06:54 -0600 (MDT) Received: from rms by fencepost.gnu.org with local (Exim 4.34) id 1GVvxx-00033d-VQ; Fri, 06 Oct 2006 16:06:42 -0400 Content-Type: text/plain; charset=ISO-8859-15 From: Richard Stallman To: jg@laptop.org CC: deraadt@cvs.openbsd.org, jsg@goblin.cx, dcbw@redhat.com, mtosatti@redhat.com In-reply-to: <1160088294.6075.40.camel@localhost.localdomain> (message from Jim Gettys on Thu, 05 Oct 2006 18:44:53 -0400) Subject: Re: Marvell 88W8388 documentation Reply-to: rms@gnu.org References: <200610041838.k94IcY6j006189@cvs.openbsd.org> <1159989002.6029.710.camel@localhost.localdomain> <1160088294.6075.40.camel@localhost.localdomain> Message-Id: Date: Fri, 06 Oct 2006 16:06:41 -0400 Most of the kids in the developing world have no electricity, much less money for access points. The only way we have to solve the last mile problem at all is via a mesh network. There has to be a powered network hub somewhere in the village, or the mesh network won't be able to reach the outside world. Why can't that hub do the job? We talk to all the manufacturers. We *want* them to compete for the business. And yes, we expect the next time around we'll have more than one possible choice. Have you told them clearly that you'll choose a chip which supports free software next time? That is the first step to apply some leverage for the future products. How far in the future do you think the next time around will be? Replied: Fri, 06 Oct 2006 14:12:55 -0600 Replied: "rms@gnu.org jg@laptop.org, jsg@goblin.cx, dcbw@redhat.com, mtosatti@redhat.com" Return-Path: rms@gnu.org Delivery-Date: Fri Oct 6 14:07:05 2006 Received: from fencepost.gnu.org (fencepost.gnu.org [199.232.76.164]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k96K74xv027392 for ; Fri, 6 Oct 2006 14:07:05 -0600 (MDT) Received: from rms by fencepost.gnu.org with local (Exim 4.34) id 1GVvy3-00034R-IG; Fri, 06 Oct 2006 16:06:47 -0400 Content-Type: text/plain; charset=ISO-8859-15 From: Richard Stallman To: Theo de Raadt CC: jg@laptop.org, jsg@goblin.cx, dcbw@redhat.com, mtosatti@redhat.com cc: rms@gnu.org In-reply-to: <200610052333.k95NXeAN006710@cvs.openbsd.org> (message from Theo de Raadt on Thu, 05 Oct 2006 17:33:40 -0600) Subject: Re: Marvell 88W8388 documentation Reply-to: rms@gnu.org References: <200610052333.k95NXeAN006710@cvs.openbsd.org> Message-Id: Date: Fri, 06 Oct 2006 16:06:47 -0400 > And in terms of mesh networking, the alternative was Atheros, with their > closed HAL. It is a non-starter as it cannot run without the processor > running, which would require us to run at about 5-10 times the power. For the record, we have completely reverse engineered the Atheros HAL ourselves. We've been using our own completely open code for four years. Could we organize a project to reverse-engineer the Marvell chip? Return-Path: deraadt@cvs.openbsd.org Delivery-Date: Fri Oct 6 14:44:11 2006 Received: from cvs.openbsd.org (localhost [IPv6:::1]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k96KiB4m016183 (version=TLSv1/SSLv3 cipher=DHE-DSS-AES256-SHA bits=256 verify=NO) for ; Fri, 6 Oct 2006 14:44:11 -0600 (MDT) Received: (from deraadt@localhost) by cvs.openbsd.org (8.13.6/8.12.0/Submit) id k96KiB1N001137 for deraadt; Fri, 6 Oct 2006 14:44:11 -0600 (MDT) Date: Fri, 6 Oct 2006 14:44:11 -0600 (MDT) From: Theo de Raadt Message-Id: <200610062044.k96KiB1N001137@cvs.openbsd.org> To: deraadt@cvs.openbsd.org To: Marcelo Tosatti cc: jg@laptop.org, Dan Williams , Jonathan Gray , rms@gnu.org, Chris Blizzard Subject: Re: Marvell 88W8388 documentation In-reply-to: Your message of "Fri, 06 Oct 2006 14:34:03 -0300." <20061006173402.GA4575@dmt> Date: Fri, 06 Oct 2006 14:42:07 -0600 From: Theo de Raadt > > The Marvell documentation does exist. I have been told a bit about > > how large it is. It sounds like you are making these statements > > without having access to it. > > Its rather small (its the interface to the firmware through a set of > well defined commands). > > By the way, you can easily understand the firmware interface by > looking at the Linux driver (it provides as much information as the > documentation). If it is so simple, why is Marvell firmware interface specification 142 pages long? Are you really going to condense all of that information into the driver, so that future driver maintainers can work on the driver after you and your NDA document have moved onwards? We found a copy on the net of the Marvell firmware interface specification. http://lists.infradead.org/pipermail/libertas-dev/attachments/20060614/bb59070f/attachment-0001.bin See, 142 pages. Not the largest document we've ever seen, but quite hefty. It looks like this is a leaked copy of the document that Red Hat got from Marvell. One of you guys leaked it, I suspect. Thanks, we have a copy now. Shall I tell Marvell that I found this? That said, this is only 1 of 7 documents that Marvell has for this chip. The other 6 documents are also very important. Marcelo, based on the balony you spoke above, I find it hard to trust your predictions that Marvell will open up. Replied: Fri, 06 Oct 2006 22:44:30 -0600 Replied: "jg@laptop.org rms@gnu.org, Marcelo Tosatti , dcbw@redhat.com, Replied: jsg@goblin.cx, blizzard@redhat.com" Return-Path: jg@laptop.org Delivery-Date: Fri Oct 6 20:29:15 2006 Received: from rwcrmhc11.comcast.net (rwcrmhc11.comcast.net [216.148.227.151]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k972TERi032080 for ; Fri, 6 Oct 2006 20:29:15 -0600 (MDT) Received: from [192.168.1.144] (c-24-218-178-107.hsd1.ma.comcast.net[24.218.178.107]) by comcast.net (rwcrmhc11) with SMTP id <20061007012833m11000d2uce>; Sat, 7 Oct 2006 01:28:35 +0000 Subject: Re: Marvell 88W8388 documentation From: Jim Gettys Reply-To: jg@laptop.org To: rms@gnu.org Cc: Marcelo Tosatti , deraadt@cvs.openbsd.org, dcbw@redhat.com, jsg@goblin.cx, blizzard@redhat.com In-Reply-To: References: <1160089598.6075.63.camel@localhost.localdomain> <200610052358.k95NwHB2006381@cvs.openbsd.org> <20061006173402.GA4575@dmt> Content-Type: text/plain Organization: OLPC Date: Fri, 06 Oct 2006 21:28:32 -0400 Message-Id: <1160184512.22956.146.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Evolution 2.6.1 Content-Transfer-Encoding: 7bit On Fri, 2006-10-06 at 21:07 -0400, Richard Stallman wrote: > Everybody in the OLPC project, including the people who decided to pick > a piece of hardware able to keep up with cost/power constraints, have > the best intentions with reference to using _only_ free software in the > machines. > > I am sure that everyone here has that intention, but that isn't the > crucial question. The crucial question is whether people are pushing > hard enough to actually attain that goal. Yes, we are. > > By the way, you can easily understand the firmware interface by > looking at the Linux driver (it provides as much information as the > documentation). > > The code of that driver would be fine documentation if it is released. The driver is in our git pool, awaiting upstream merge at kernel.org. > But I think people stated here that Marvell has not released it, and > that Marvell says it can't release that code because it is too tangled > up with the microkernel. Marvell can't release the firmware that runs inside the wireless chip. The Linux kernel driver for the Marvell wireless chip is already available, and has been for several months. http://dev.laptop.org/git?p=olpc-2.6;a=summary > If that is not right, would someone please > set me straight? > > > The Marvell documentation does exist. I have been told a bit about > > how large it is. It sounds like you are making these statements > > without having access to it. > > Its rather small (its the interface to the firmware through a set of > well defined commands). > > I think we are having a miscommunication. The documentation we need > is the specifications for the hardware itself. With that, we could > write a driver. We do not need documention of their driver as such. Please look at the libertas driver in the git pool. Regards, - Jim Gettys -- Jim Gettys One Laptop Per Child Replied: Fri, 06 Oct 2006 23:06:18 -0600 Replied: "jg@laptop.org rms@gnu.org, jsg@goblin.cx, dcbw@redhat.com, mtosatti@redhat.com" Return-Path: jg@laptop.org Delivery-Date: Fri Oct 6 22:41:28 2006 Received: from rwcrmhc13.comcast.net (rwcrmhc13.comcast.net [216.148.227.153]) by cvs.openbsd.org (8.13.6/8.12.1) with ESMTP id k974fQKW026611 for ; Fri, 6 Oct 2006 22:41:27 -0600 (MDT) Received: from [192.168.1.144] (c-24-218-178-107.hsd1.ma.comcast.net[24.218.178.107]) by comcast.net (rwcrmhc13) with SMTP id <20061007044116m1300ngqa0e>; Sat, 7 Oct 2006 04:41:17 +0000 Subject: Re: Marvell 88W8388 documentation From: Jim Gettys Reply-To: jg@laptop.org To: rms@gnu.org Cc: deraadt@cvs.openbsd.org, jsg@goblin.cx, dcbw@redhat.com, mtosatti@redhat.com In-Reply-To: References: <200610041838.k94IcY6j006189@cvs.openbsd.org> <1159989002.6029.710.camel@localhost.localdomain> <1160088294.6075.40.camel@localhost.localdomain> Content-Type: multipart/mixed; boundary="=-mzrL4m/DzYrsoEs6bqCp" Organization: OLPC Date: Sat, 07 Oct 2006 00:41:14 -0400 Message-Id: <1160196075.22956.232.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Evolution 2.6.1 --=-mzrL4m/DzYrsoEs6bqCp Content-Type: text/plain Content-Transfer-Encoding: 7bit On Fri, 2006-10-06 at 16:06 -0400, Richard Stallman wrote: > Most of the kids in the developing world have no electricity, much less > money for access points. The only way we have to solve the last mile > problem at all is via a mesh network. > > There has to be a powered network hub somewhere in the village, or the > mesh network won't be able to reach the outside world. Why can't that > hub do the job? A single access point can't reach far enough. Despite us making the 802.11 work better than anyone else's, access points can't go far enough, nor is there always power where you need it. And there is no guarantee that power there will be reliable. Hell, Michailis brought back pictures of a school in the capitol city of Nigeria: yes, there is power in one room of the school: none in any of the class rooms; light entirely from the windows (see attached). It looks like a dirty version of a 19th century 1 room school house in the U.S. Outside the capitol it goes downhill from there: 30-40% of schools have no power at all. Well over half the kids have no electricity at home at all. And Nigeria is a rich country on Africa's scale. Our focus on low power and mesh networking is due to the realities of most of the world, that few of us rich Americans ever see. To serve the kids, we have no choice but to design to the real world conditions. > > We talk to all the manufacturers. We *want* them to compete for the > business. And yes, we expect the next time around we'll have more than > one possible choice. > > Have you told them clearly that you'll choose a chip which supports > free software next time? That is the first step to apply some leverage > for the future products. We make it clear with the manufacturers we talk to. It is an interesting cultural exchange at times :-). In general, we do not lack leverage at all; manufacturers are very happy to do what we want, *so long as it is within their power and ability to do so and they believe can make money doing so*. A good example is Marvell's (yes, the same Marvell company) reengineering of the SD interface to solve the documentation and driver problem posed by the SD association's draconian NDA terms. They revised their SD hardware design to conform to the public SD spec at significant expense; they could do it for us, so they did it for us after only mild arm twisting. The wireless firmware situation was/is different: Marvell doesn't own all the IP and could not do what we wanted in any legal fashion and there is no other vendor that (yet) makes a suitable part to use. So, plan B: replace the firmware entirely via Meraki; but that will take longer than we can wait to start testing, or maybe even early deployment. Time will tell. And, of course, we tell all the other vendors that want the business that they need to make a part similar to what Marvell has done if they want the business in gen2. It was lots of fun working with the touch pad vendor: it took about 3 times to convince them that: 1) we couldn't care less about Windows compatibility, 2) we wanted both modes of the data from the device at once, (it is a novel dual mode device; a cross between a tablet and touchpad) and 3) we could make both our operating system and the window system use a device that didn't conform to "Windows standards" without difficulty, since we had all the code. Once they understood we were serious and did not have Windows constraints, the answer to "jump" was "how high".... Bridging the cultural barriers can take work, however. > > How far in the future do you think the next time around will be? As soon as we get this one done. Of course, we tell every vendor that comes through the door what they have to do if they want the Gen 2 business. Very advanced planning for the 2nd generation system will start before this year is over. We should be done with most hardware testing by sometime in 2nd quarter of 2007, and that's when we'll start in on gen2 design more seriously. - Jim > -- Jim Gettys One Laptop Per Child --=-mzrL4m/DzYrsoEs6bqCp Content-Disposition: attachment; filename=Classroom.jpg Content-Type: image/jpeg; name=Classroom.jpg Content-Transfer-Encoding: base64 /9j/4AAQSkZJRgABAQEASABIAAD/4ZRaRXhpZgAATU0AKgAAAAgACwEPAAIAAAAYAAAAkgEQAAIA AAAMAAAAqgESAAMAAAABAAEAAAEaAAUAAAABAAAAtgEbAAUAAAABAAAAvgEoAAMAAAABAAIAAAEx AAIAAAASAAAAxgEyAAIAAAAUAAAA2AITAAMAAAABAAEAAIdpAAQAAAABAAABGsSlAAcAAAAuAAAA 7AAAe5RDQVNJTyBDT01QVVRFUiBDTy4sTFRELgBFWC1aODUwICAgIAAAAABIAAAAAQAAAEgAAAAB MS4wMCAgICAgICAgICAgICAAMjAwNjowOToxNSAxMzo0NToxOABQcmludElNADAzMDAAAAAFAAEA FgAWAAIBAAAAAQAFAAAAAQEBAAAAARCEAAAAAB+CmgAFAAAAAQAAApSCnQAFAAAAAQAAApyIIgAD AAAAAQACAACQAAAHAAAABDAyMjGQAwACAAAAFAAAAqSQBAACAAAAFAAAAriRAQAHAAAABAECAwCR AgAFAAAAAQAAAsySBAAKAAAAAQAAAtSSBQAFAAAAAQAAAtySBwADAAAAAQAFAACSCAADAAAAAQAA AACSCQADAAAAAQAZAACSCgAFAAAAAQAAAuSSfAAHAAB4ggAAAuygAAAHAAAABDAxMDCgAQADAAAA AQABAACgAgAEAAAAAQAADMCgAwAEAAAAAQAACZCgBQAEAAAAAQAAe3ajAAAHAAAAAQMAAACkAQAD AAAAAQAAAACkAgADAAAAAQAAAACkAwADAAAAAQAAAACkBAAFAAAAAQAAe26kBQADAAAAAQAmAACk BgADAAAAAQAAAACkBwADAAAAAQACAACkCAADAAAAAQAAAACkCQADAAAAAQAAAACkCgADAAAAAQAA AAAAAAAAAAAAAQAAADwAAAAcAAAACjIwMDY6MDk6MTUgMTM6NDU6MTgAMjAwNjowOToxNSAxMzo0 NToxOAABcMAAAHnsAAAAAAAAAAADAAAAHgAAAAoAAAMWAAAAZFFWQwAAAABPAAIAAwAAAAIBQADw AAMABAAAAAEAAHKsAAQABAAAAAEAAAjKIAAABwAAcqwAAAjKIAEAAgAAABIAAAa0IAIAAgAAABQA AAbGIAMABwAAAAgAAAbaIAQABwAAAAT+//j/IAUAAwAAAAEAAAAAIBEAAwAAAAIBjgF6IBIAAwAA AAEADAAAIBMAAwAAAAEAAgAAIBQABwAAACQAAAbiICEAAwAAAAQAAAcGICIABAAAAAEAAAr0ICMA AwAAAAEABQAAIDEABwAAAAIADAAAIDIABwAAAAIAZAAAIDMAAwAAAAEAAQAAIDQAAwAAAAEDvgAA IDUAAwAAAAEAZAAAIDYABAAAAAG4iYjwIDcABAAAAAFAAI9gIDgABAAAAAEAAAAAIDkAAwAAAAEA AAAAIEEAAwAAAAEEpQAAIEIAAwAAAAEEoQAAIEMAAwAAAAEACAAAIEQAAwAAAAEAAAAAIEUAAwAA AAEABAAAIEYAAwAAAAEgSQAAIEcABAAAAAEADwAAIEgAAwAAAAEBLAAAIEkAAwAAAAEEpQAAIEoA AwAAAAEABAAAIFEABAAAAAEAAAADIFIABAAAAAEAAAAAIFMABAAAAAEAAAAAIFQABAAAAAEAAAAA IFUABwAAAAEAAAAAIFYAAQAAAAEAAAAAIFcABwAAAAEAAAAAIFgAAwAAAAEAAAAAIFkAAwAAAAEA AAAAIFoABwAAAaQAAAcOMAAAAwAAAAEAAgAAMAEAAwAAAAEAAQAAMAIAAwAAAAEAAgAAMAMAAwAA AAEAAwAAMAUAAwAAAAIAAAABMAYAAgAAABgAAAiyMAcAAwAAAAEAAAAAMAgAAwAAAAEAAQAAMAkA AwAAAAEAAQAAMBEABwAAAAIAAAAAMBIABwAAAAIAAAADMBMABwAAAAIAAAAAMBQAAwAAAAEAyAAA MBUAAwAAAAEAAAAAMBYAAwAAAAEAAAAAMBcAAwAAAAEAAAAAMBgAAwAAAAEAAAAAMBkAAwAAAAEA AQAAMBoAAwAAAAEAAAAAMBsAAwAAAAEAAAAAMBwAAwAAAAEAAAAAMB0AAwAAAAIAAAAAMB4AAwAA AAEAAAAAMB8AAwAAAAEAAAAAMCAAAwAAAAEAAAAAMCEAAwAAAAIAAAAAMCIAAwAAAAIAAAAAMCMA AwAAAAEAAAAAMCQAAwAAAAEAAQAAMCUAAwAAAAEBkAAAMCYAAwAAAAEBAAAAMCcAAwAAAAEAAwAA MCgAAwAAAAEAAAAAMCkAAwAAAAEAAAAAAAAAADA2MDIAADEzMTAAADQyAAAAAAAAAAAAAAAAAAAA AAAAAAAAAAAAAAAAAAAAAABJBcTAAAAAxgApDsAA7gLbAAAKbzEAAAAAAAAAAAAAAAAAAAD///// /////wD/7ACNAQEKAABkvgAAR3z/6VD/6lP/7W//86T//OQABf4ADuYAF6QAIFAAKQAAAAAAAAAA AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAExvbmRvbgAAAAAAAAAAAAAAAAAAAAAAAP/Y/9sAhAAB AQECAQECAgECAgICAgMFAwMDAwMGBAUEBQcHCAgHBwcHCAkMCggJCwkHBwoOCgsMDQ0NDQgKDxAO DQ8MDQ0NAQECAgMCAwUDAwULCAYICwsLCwsLCwsLDQ0NDQ0NDQ0NDQ0NDQ0NDQ0NDQ0NDQ0NDQ0N DQ0NDQ0NDQ0NDQ0NDQ3/wAARCADwAUADASIAAhEBAxEB/8QBogAAAQUBAQEBAQEAAAAAAAAAAAEC AwQFBgcICQoLEAACAQMDAgQDBQUEBAAAAX0BAgMABBEFEiExQQYTUWEHInEUMoGRoQgjQrHBFVLR 8CQzYnKCCQoWFxgZGiUmJygpKjQ1Njc4OTpDREVGR0hJSlNUVVZXWFlaY2RlZmdoaWpzdHV2d3h5 eoOEhYaHiImKkpOUlZaXmJmaoqOkpaanqKmqsrO0tba3uLm6wsPExcbHyMnK0tPU1dbX2Nna4eLj 5OXm5+jp6vHy8/T19vf4+foBAAMBAQEBAQEBAQEAAAAAAAABAgMEBQYHCAkKCxEAAgECBAQDBAcF BAQAAQJ3AAECAxEEBSExBhJBUQdhcRMiMoEIFEKRobHBCSMzUvAVYnLRChYkNOEl8RcYGRomJygp KjU2Nzg5OkNERUZHSElKU1RVVldYWVpjZGVmZ2hpanN0dXZ3eHl6goOEhYaHiImKkpOUlZaXmJma oqOkpaanqKmqsrO0tba3uLm6wsPExcbHyMnK0tPU1dbX2Nna4uPk5ebn6Onq8vP09fb3+Pn6/9oA DAMBAAIRAxEAPwDvLa8sJfuXFx+MI/8Ai66R7WytLZZp744ONsa22Hb6HJx9a8zi0O806QLDbSSy v0YqOMegqf8AsrV2JMlpd5P1rr9vBnZ7KVz2WLxDbS2vlW9wYYUb7hgZdx9Se/1qGPUn+X7NcwlW 4+YsMfXivGLu01m2iDJaTAdy4brVK307xFf3QS2IhXGQsikEj27n8qtVoGPspHvUuoTrkS3umcdv OYH/ANBrMk/tG+QvYvYvEmSzidQFHvkiuAHh/UIkG6wu7p8ctJlF/BRyayp9I1p8m53LCcBIvK+V cemBV+1iCpSPW9NtEWBGvLyO6c8hEuokX8ctuP6Vq3ep6g0QjtSscAORHHNFgf8Aj3NeDR6XqjSB IQNw5OIyv9Ke+j3UT51CWR8fwxKW/U4Fae0i9CXTkenXKajeFk8q7lB6qroR+hNUbfQNUgcGKwuI xJxvlJVc/UkCvPBf/wBnhjBb7WH8cpLkfnwPyrNvfEElxIDNcSu44GeaftYmfs5Hutv4dubYE6zJ NJjqlvbMfw3HgfrUn/CQppDYsrKS2Kg4kMZZz+JHH4AV4Ot3fOu5cQK3O+Z9nHr6mrFtcQ2zA3Fz cXUg52xuY0/PqfyqnWiiHTkem3fjImctm6lkbncQWJpI9S1TUU3QW8kUTcebcHYv68n8Aa4dvHt5 bRbNMkNoh42wkhj7Fjkn86gs7vVrkmSBroLnJeSQqM+pJNL2q6h7NnfBYUctrd9PO3TyrZNgP/A2 B/8AQaki8WRaQSdGsI4nXjf5bSv+ZHH4VyC31xGo/tDVZyc52W7k/mxP8qtJ46m0+NktLiVF6Fml aRvzPT8AKftEPkex0l34g1vXXMiRzkN/y1mbYP8AvpjWcNPZ2DatqCbv7sKbz+ZwKoweLtUviWgk uJQeSzJkA/UipYtZuZG3axc28gP8DQrIf0GP1qXViP2b7HR299Yaeny2ktwy/wAVzLuGfZVwPwOa 5fXPH2qXBVNLlNgq8BLdAgZfQleTUs+tWhRkSysm7bmiUH/x3H86qweIobaMIttbqegKbo8/kaxn WUtLmkaT3sef6p4d1zXAHQyQJjrMdgPvycn8M0af8OG663qUPBzthiDfqx/XmvTbe4GoLuXTmcHq 7SyKv5lqvxWtj5xFzbwYAz+7lkc/q2K8upXpU/iZ3wpTlokYGnRWmjAiz3s7DBYy5J+oUAfpW/Dr 9zKRHaCYhT8qoD1qX7LZmfNraHauB88jEenTPNdfYjaiRwlYlJPEa4FeFUzenDSmr/gd8MulLWbs ZsevawmFlkuIiOcNK2fyH8jitqHUL+RB9supnU8fePH55/Litu0tIlIypkJ6k1vRiNUACqF+leFU zavL4dD0oYKjDdXKFsbudBlBtPJ/djk/U810lrYzNGscr7VXtnj8BSWtyu0MCeK1I5WL7osN9O1e DUrVa3xybR6KjGn8MbF6HTIVPO58epq+NkBHkRjd2wKgt7lJMLGfnPGPStlYBjPT61yWKcnfUhNz sKiQc9z0q75+7AwDn1qOKwXfvkfecevFO0yeKaZ+TtU4zjFVbsZ3NeKAbOOBXKeOIBJ4bfA+46nm umbU1aTamdq1i+K3W+8M3RXOIyn8xW0bXROp8xXzG1f5s4Y1QiIkBC/StzVLXzIiB65H1rEjkVZM rwK6pqyOmB1nhFN8NwigZDDP5V1aWxSUZxjrXLeDLtf7RuI0A+YA13MgMcgaVTgnsOa859TqkQzW geEkAg5rm7+32sBtJH0rs1maVyphdRjqa5DWrqWEjyLeWU5x8orGK2FqeVRWwY8VqRWO5OeD06VK ieXIA3U1rRR7lyM10NsjqZg03avU/kKWO0G7J6D2rdVBKDwRtOD2prW2G4GKnmE+5R8o4wDn8KWK Bo8mKRl9hxV9oSFbZjcAcZ9agSIqieYQXAAYgcZ70eQ2wkkl6ebIxHP3iajkaduGmkI92NWFUMSS MYqKRSeMkY9KXM0ZlGdJdoPmFl6YPaoJo/OjxIFI9lArSeRWXI6VXZQE5zirc5LqOyfQzvs4lTaQ MjPVfXrTVthjYyoyjj5lFaUcYSQ46HrTFHZWJIGMnvV+2qLaT+8lRjfVfgZ4tYklUxQxKynORGBn 9KqXFnFcNm4ijc+pUGtPJLHb1xWc9x85AGDnFarEVV9p/eS6cP5V9xi3Ph+3kADxLkdSAB/Km2+j QWD5ggj59QTz+eRXRSfOMvn8KzjeRFyu/kVaxWIe0mN0qfYyrzT2lbkjjgYJ4/M1kPpluhxMLg+/ m/8A1q6qeURYD8g8ZHNMa0ZkJUbgB/n+YqvreIW0mUqFLscmNKtWDMiynZztLgfnxToL1bZSNOSG Jv8AZXnH1OTWf4c8RQ+MPCtvqFlFNBFdoTtkGGGGIwfxBrWitVU5xjFclTFVpO0pM6o0acVdRM+Z priQtLKdy9yc10FoEgjDDILdeKIrFGcZHHJ5rbEUW0Aj7vasOYTQWLbgwbGT7V1GnwFp0YnCnt3q jaJHFGdoA7n6Vo6dqkV7GstnIjx5IBHQ4ppN6mUn0O5iiKlBHjB6n0FaH2KIuXmJAUZrAstSgjK/ apQNo3MO4FMfxLbajIILWOeQSHAYDgVtZ3OI6ZpYAQxPyqOMVoWeoRA5JAXpiuTSPyXB2qwHG2ul iUxYIVF3HnIzx7V0xppGTkdTC0cjh40yx4GKW/uWkcRIwTueev4Vnq6Q2hlRisikjAz0qNNTtriY btu49WHUiokktUJMRtPv5wRayn5u4B6e9dYIlhtVjuQylAN7gcGrVpqUEfG7CjrjrTWmDowkXcrH OTjpWCV1cq5mRSRzu/lO7Y9EJPpxitvWLKOz8LXUa/Mzjcfw5qaWJLKyWRlVm3fKBngep96zdZhk n0q5dTkiF/lP0q5Lkdk7hB8x81394GdlZcBcmuZkVUmBwfm+bGev0rdkkX7S3nL94cE9xWVLGvlK RwQfyFejNcyNYvU7PwaqC/keJDuaPn869HijYSjzccelea+CLjZeyq33toOfx/8Ar16L54N4scYY 92JryJOzaO9mjMiuS2cKvPpWLd2wkJ3E7B1wc10k7Rqm1wCSOlYCjZFIWA24z171jEi588XLrsG9 25yBnrmtrw7fm5zFISXUYHuK4i91+ytrcyXVzEoTks0gAqlZeM9JMCSWuoWih+n75VJ/DOa+rlSU o8rR40ajTvc9q8xvLJjXnOBkVBd3i20YecMQnJAHU18l/G39svw1+z/qdtpnie31O4vLq2+0J5EQ ddu4ryS3XIrwC6/4KgeB7mJh/ZPiFtuMkRR4J/769a44YCtUSlGOhvPFQj7rep9d/ETx/q1hDcQ6 Pbm3UYJuB8xwemOwr5+u72/uSn9oX91KzhXA3nAyM815ZqP/AAUt8AalpdxDcad4g3OhRcwJ17fx duK8m/4bT8BXTj7P/wAJB5aqoAFmM8cf3q+ywVKVKDjKnZ+h83iainK6ldH2toV3eWFwjwahdMUI +QSnH4g17nBe3mpRJNe3Ei7hnCjAxX56WP7d/wAOtI0029no3iV5X/1kxtF3N6dX4rOf/got4f0y JVh03XjCpHL2yqcd/wCI1VShKs3+7tbuiI1FS05j9IhNcK4a2mkVsjPGeK3Y9TnY8FcKMk7a/NJf +CmHgxQCuleJHJA+VYYh/N6qL/wU28HuzRw6P4nVXOOYouD+D14s8FUk/g/A9COJjFfEfp+NQni7 hvYDNWV1QpF+/Tc/T5Qa/MCH/gpl4MtiWnsvEYYDp9nj/wDi61dM/wCCkHgvUlJZtVizx+9hjU/+ h1k8um3blLWLUdbn6My6sincA5Ye2Mmqq6rH5hwhYH0Nfm9df8FJPBa3AE8esugGP3dujH/0OqkX /BSD4esQIYddUg/xW6j8vmrB5fKP2WaLGJ9T9Mb6+CWJa33N2C96w7e187JUZI654P8AnrXyF8J/ 20PCfxW8Z2mh+GBqour7f5RmtwqZRS55z1wpr7ftraKVwtw4QEZ6VxunLDPlas38jujUjWV77Gbb QbQ7ROMI20gnOTWsj+UG8qXBAx8p61Za1htZAEcbTHuBAzzWeNQiWZkZSxUkYArlqTlLR7nZCCRy GoT3SyEFFVSoCkfjWTZ6p55dLYhmjOHBODng/wAq9AkQ3dkXCEKpweK5mC0jijEkqBn5HTFeLdPR nrdC7ayvFADIWZ8Z6c1oWNwzOTdKEBHHPP41lXMzgKIRkc9vSpYbeW7gmLKRsXue9dKV+hzSdjN8 R3dxeTi206fY04Ccccd62dE0C30S3jR2klEWBjseaj0e3VYCwAklVyTk847Y/A12Mca782oKEdc9 69N2jFRPPV27naQ3m618wAAt8uMdasadZEgS7kXbwoXr+VU4rHNvEQSf731roIAYVjdcbW4+uKVN Rs2ckpO9i1aWaRsrTvhuoBrbdlwqu5U9BxWSXjZkEufmOMCrMkyyMgiAJPANdMmmjCN0Zt3LP5AD JIIZCdjKMbgOufzrDs7qZ7kQ2qne3LsV4UfWu4tLZvPHn7SzH5VHp9K7KHSoIyMJwB8xA7158nZH UnY8jgk1B77YoYxp8u71+lez6amy2RJgemWyO/HFVIrTbOFCg8da6e2swJGy3Q/KPb/OKTjpcHIu LHuBMnykcBT2qPUgselybRwqMvJ9sUqujP8AK2/HPT0qprUix2MqyHO7qPyrna0YovVHx6Jw6OW5 IbAJ9KsmT7oVRyMk47CsjSLlbm4cYG4HleuT0P5V0M0bQIGXnA/LNdzl7uh1qNpFzwzKthrZeTO2 SMjGO+RXfWuqC6kyNyBuoIry3SpftmppFGCzKCcivTbWwul4ljVEk6kda8mT1O+2h0O0SugVsgAV Nc2paIDpn0qoyx20iLGCgTAAIxwKuRTg5aTARunvRojmP5HbnxddX0Qnla0iaMbvKBkXJzjoD16c 1xN98Tr24ci4sdPCLznL8kf8DrvNJsrySzlTSp7dWkVo/wDU+Y7jGSAcE9u3rXJy6TJd7DdxW0gY HD4wcjrnHWv6LdPqfmSlYhuvEc/xXltRr1vEjWaNDGsZY5B5ySxJJ4qeLTIryKAWQKBEUt5eW3Ho MgHFaPgTSha+KbXzvumVQEH8WWxg/gTXpeh6cNK1O5gZIo7cOSoGCQc/KOOa6acLJIynPmZ5DF4P ivp5J/EVhLEoVgr7OCy8f/r/AJ1o+CPBDi8uIbeQCFolnRdoJ5yDjHbI69K+gPGemQan4cjnMkUP lgsUx1LcH8ea820HztFu1nt3UvGgjBJz7/zrfljGWpz3ckWz8LLmVIZLWGJ2yTjoTz359K5DxV4C udJu7dLs7VJ/1e/eGGc8c8YzXrFnr93YztPbFixO4kttxnuD69ao+LfFX9o3cJu4WdsR4ZSCVx15 7546VrJRsZLmOMk+HsF/fQvaeRbhANrhMBvQn8ulcv458FaXoHiKDTzLcNNLGs6uoVV+YA8juM8V 6vr3iddNjtoNLQDzYxI6bu+MV5TfiPXblZ9RjIul/dgk4AUen60pW2QJvqeV6p4fjheQTSCQrwux SfpUehfC1tcwDbyRxllyyrzz6V65F4XWwv1aZvLt7kgRs3zDJ9SemMjrV0wT6J5ixXwngSTdkEnO 09MDpmud04rdG3M+hjz/AAKsfCl3D9pvlljkTLh1+ZM+3aq1x8HtI1CyMthcpPMxI8reEYAd/pT5 tSjt9PmnvxG9xcptG6MnAOCNp6j0rltLur2XU1ldGWNGHy42gcHI4/zzS5Y7WEpSvua/wIu7n4G/ HfTr4xCeO0uRH82VBEqtGfx2uT+Ff03M261YoN0u35Bnviv5mvGMdzpjx3WpRt5bMjIC5PKBvyyM V/TTpyCeCOYBj5kYYZ9MV+d51T5XGS8z6nATbTRjGds4kV1YHb1yOQDW7DZvLGp8thtHJP8AjVuK 2WZmC/eBweKx9VC6dcxmMERuBu7dfb8a+B0l0PqotrW5qx2+LYs5jxu8sruBI79PTiuUkkKyhRjh se1bw0+STaI5BH/EQ5AJrmtMUx3Sx3UizSRjMkmNoJA5OO3fiuT2Lk1pa53+0STNhgs042uQwU/d 6dKv2QaKC4Usx83r+X/6qbZTqD2y3txiuk0uASSOZSANvHFezTw0X1PHq12lsczp9nIC7J91ByD7 VcXUvsfmebGX3HcMHFTrpVvLeRyl3MkZK8E7e/asO8vRDdEZztBzxnGK6K2HjFaHLTrSk9Tqbzxy umeHprwWs0scEbySBWAIVRk9faviXxR/wUP0awwNC0+eXyzj946/Nx254r7Djt01bwXqsCldzwum R0yyY/rX83Oq2qysSFG4cZxTwGHjiedS6HxOfZlVy5wdLrc/STUv+Cjl5KX+w6ZAgB43S/hXF6z+ 3p4yl/faJLZ2yBTyYzIcZxx2647V8MXfgHV7HR01DUNL1C3sZdoS4ltnSNsjjDEYOQciks4Ge1JQ t02cH3H+FfVQwFBaWPyifEGMqX5p/dZH07qH7aPj7Vn3y69Op6fuYQn64r7e/wCCfXxw1/xn428S 2/ie/ub5fsUcyLM+7BD4JH/fVfnj8Gv2efEXxvnvV8AQWkq6eENw891Hbqm8kLy5HXBr6m/Y30q/ +Fv7Q3iDRNeESXdvaTWM6pIrqJEmjBwy5BHBwRWeOwtKlQm4pXRrk2ZYmvjqUalRtX7vsz97bOGS 5sVm2nYjBGI7nr/j+VdBHMqLiLO7kt6D0FcHpepvcaX+6kym4MBnjP8A+rNPttQM13hCUKj5h+tf Ac8baH9F8rvqdVpd3KJZTdBVXYQhUZ5xkH8xisLxDA09u8ksu0DBIHTHP+fwq2twIiACQoNc7qur btOIRwFZSTgdFFZVJaaFU4+8fKVtcmLW7whgFS5kVcYxtyf8K760k+02TvE+5VG1uwP+ea8c8OTh /FesRSPnbdHqP4Tz/LFerNOE3pbEAKuDWkV7p6Mn7xHp050rWEMnKknkDpweK9rjuWaBneTKtgRp XhSF5bmL5Qd8gGffp/jXuLRoulIGUhwnQ15r+I7G7xNWeQLbReacsBz6YqrLcIbF0A+cDNYOnzCZ o95Zs9uuPrV6+sGjG23chmO3GM8V0KHM7I4b23P5cdDu7eS1MdwxRfm4CcjI9R64H60+7sLCbYGe aBy+zbHHkDkevselZ+qWMVjaTGzcKp4bEmeMe/am2umwtbBpJlVydjbuuT3z+Ff0hex+YMwmaKy8 QWzWgd0tZVbeygb8MPQ+tehanH9i8R340pXjgM7PDIc42ZJGDXnt1AbDUXETxSbGESlOhH3enTPy 9a9OuLl720tJFiLRJAkbg5GCqgH9f50RegmtmW7mK3l0yBDPHuLMzKvJ/Kt3SfDtvpmiNcXcJZ25 68rwSD+e3Nctp11CdUgbZHDAp5AGQce1dxbX21bltsl4jMWDH5QO3APXr+laqSepi00rFb+09Mu4 XguIIhIuQpjUH6fyrMbw/BfWRbTrtGlDE+U67cj2x9BVfTPCMlxrO+GO5VJSwVmGOecn88UeI/Cv 9juAZjE68+wwCD75OP1q9Xqw0WiOAvvD6JGlzd3VsZs4KZIz2x/SsyPRn2tslibzwC4Dgg9f8/hU 8+gQyX20XHmkk8FiBzV7/hETbptWR1miII+YkY78VC72KvYyLDw7t1B7a5sp5IpYwePugZ6jPaod Wezm3jy1LISrKwPWvQJ9NaEwz28kizhNrqXOSfX6VzA0LN1LI05LSjcV25HfNElYhO7OMWzt4beE hwiyRbchd2Dj/ECt2xmt9MtxHbxvM0mBueMgZ6EA49qp6poO1PLgmACx+ax242jHT34NZeiatBDY CGKeSfyHLFWONv0qNimXPHNpc3/hY3GqKQis+cEcsAf8DX9Fvw81n+1fA+i3c2wNcWEcrAepUHiv 5tPGXit7vw7KtvIr26rvVCcsN2VPUdec1+9fwA1N7n4L+EZpCCW0a1kyPUxKTXx+bNWTSPYwfVH0 e/yTNsjxu5xWVreopCwBbcwPTFVF1JzbqVcoSv6VwvidZmmFxHIFVcFwW9jXxsJcr0PffvaPY1Fu TK8roy7Sw5FZEcwbXJkiclWYHHTrXAaVdTbSbtydzFgenGeBj6VpyyPbXe4n5sYOD29a8WtdSU2j 2abXLyo9ktSBkEAqPT+ldHa3AEKJ9wbCo75rzzTbwEHDgkKMZq3f63G+lu0D5aEEcHHIrppVbHHU p3N3UryOwhDTybdz9+OTXn/iDWYrWdiGDNIOSD6V5z4u8Sz6hpcMSPKJJZgwweg/yK8tu7+5V/3s u9lG1ssenr+ld9SpGS3OaEXc+oPBniOIC7jYHy32HBOeor8CdXia11O6hmOGindPyJH9K/XPwP4j VdaZLx/vohKqfTcK/KD4mQG18davDEeIr2ZcfR25rpyuKUpv0PzLixPlpvtf9D9S/iT8SdI1r/gn 3BpE19oUGoW9lYLHZpeTRzysjoGYwkmMnGSWGOMnHavya0+4AsgNqgvJjAPThjXPsxSJmb8c96ta VLub5SPlJwCfbFfW0oqGh+IxvKO2x+gP7EXxxsfgh4u1S58S6zdaTY3sKxy+U9wvmgbuCIVYkAnu DXR/Dnx7Y+IP2v8AxFquhStdWWpy3U0EnzDcrSKwPzqp6eqg+1fn3HcZDAD5vp0r2b9nrVhpfxXs pZCUTbIpJHQEVyY+mnRqNdUe3kMnHHUbdZI/oi0vxAttp/lrJ/rRlTj2q/BetB9nmRHkaTbG6ZwR zgmvmHT/ABtFbWaSXLAgJuPpgDj9K7mL4l2zWo+xybpxnB+nNfkUGo6M/rmUb9D6O1C7+xaf5jlW c8EZ7V5ZrHiFb3TP9Eby2IIyOe/p6V4rN8TXu79AzzRAfLIzjqR6D0xXPar4q3SvHBJyp4w20E46 Y9sVbd7oIxa1ZzMF+i+OtQMb7N5UEA/e+Vf8/jXqkOuA2TyIuAWx9c5/wr5A1rWZYvHs74KJdRwS KQcEt8qkfmDXt8Opx2ulJJOcQycAdTuHX+dd0IuyLmtT2Tw7qT3uoxByNpkUYHODn/69fQmr3KpE piOFRQrCvhzwb4mbUNZjhsm2fvlPI5ABr6w1TxBAkTrcyKgXnn0rjdvaXZbdonZ6XKI0IXBVeu45 4rnr3U1aeSd3+fPGM9q4uLxtBBeOiSqYyuQRzn0rmF8ZRT3RjjdBubkt0qr9jK3c/niis/tGluXD uVJy2cKRxwf0/MUlvbINOInQndyrY5HP8s5rY0+xks4I3vrdkWd2AABUEjHQ+3ymr01sbedl0aOQ wLGHZmX7u4jP5Gv6HbSPzNI85u7V4rlGjUIqSeaTwGzx36np0Fd5pkL6hpAlndkijkzjHGMgfmK5 2/tpWigPmI2cjGfunb3H4/pXpHhWLzNHWzvXPzPls/iaiDuwmrI1baGPTCFt1juznJYgjp7YrsHu rrVNHBnaIRqSqxoNoHvkD15/CsdonsoomkC73ONrccHNb2nRSmzMIIyGyFJwc8//ABVbRZnNHI2K X1vfI8TMzgccg8Y9+1dX9p/tezjtrwrEANr5YEnkgkfrV9p3tZHeCOJfssW1yo3DB6nn2rzq41NL e/kmiRBbxjc7hjlfTJ9eRx0rTn5TLluyrP4Wh0uYtHMqlZCRuTcWHpWTqNiywFwZGKBgSGJDHn+X 9KZN4yXUreaKGNZI0wFmZcE/l7ZrJfXVeMiRZCzcFR2HGapVFbQXKyA3N/8AYPk8xABlRGCvAPf1 qbS/FF5qCSCOMRXkagZdMKQO3NR3WsySWUbWYkCoeUKbQc5z179a4258R3OlyS/2k1uIHzt3zIrJ x6Z6Zo5kupVr6DNV1G91e5KBkSSJTG28lfzx/npXFadbvYlbRY1EzkvIxx24yKu3fiCwvrkNcXMa M3LCL5wMf7uT1rHt47q7vpDa21/dBxuBjt5G7cdv84rBzXc0UW+hieK9QUabPGiKNkYA9yCc1+8/ 7K2tLP8AALwgWAdW05YtwPdflx/46a/Dm4+Hmt69p7JDpWsO+11DCzZVJxxyT6nn2r9W/wBnDxUf A/wc8OaDrBKX9vHIJAo+5+8bA9jhh+tfI5pO8Fy73/RntYSn7zTPt1/FltpsG66lVEXn5uuB61xn jbxVbrbiOJldJVOMHOQB/wDWr5d+Kviy506dfJLuk+cNkdD3/M4ryfXfH1xK9t5jGPy12rt4785/ M/ma+OhJztY+hdJR3Po6X4gJLe7VaPONuOa7PV/EAt47eV8HzYg3y+nTj9K/PhNc26tCI2JOclie K+stc1B9J0DQ7iMbhGArq3Ofm5H5Vz14bJ7nZTXLex9aeHj59qTAFZpBgtnvXkOt+LoNCFxYzzl5 2lKgHgEHv9Af51uBp0nhTQX2RTgFjx14HHp0xXyT4/vpL/xncTzzAHztmzHChTj+fNeSlebOlbHb XfjeG5RJZZ9hjIXgYzjiuX1LxTHOSDMUZGwV3e3H8681+wS38ca2u5lUbjkDrn0plzobtcMZUki8 pyoZSRnHr+ldFlHW4kuh6H4K8SI+ubpXAacqg564P/6q+O/i1aA/FrxEsiYzfyPwfV2P9RX0/pXh 0I1tJb5ysijI79a8D+MdusPxi1pQrZLo5J9Co5/Wvby2VpyXkflPFcb0YN9/0PHGtF8twOo9u1Q2 yiAnAJwwGSemev8AKt5bYSGTyycdAc1ltEYmIIOTInOf9of/AF6+2joz+f4vUczsAMZ5ruvhLeMn xEsByQwlyucZwjEVxE0LREhiGIOOnFb/AICHleOdMEQLM8hXgc9Dn9KyxqvQmvJ/kexlLUcZRl/e j+Z+lekeKkttONqRujlQjPOcn15+v51uxeMfItGS3ZS6Adc5x0rw+S0uIbuOVM7Xy6rg9scH0rdc SabqIivh5YmXccHkj+lfhaumf2s4rqehaz48nndpbb92QMDPA4xyfzrmn+JNzfRRtIY2aA5BQ9Dx n8gT+VZ83hufxFg2KzDEZYqDggAjOR9Oa5e70O48OlonIf5gcqM46/z5rrTfzMVFHX+KdNItpbvU LubzLSzjkRYyMMPMdjn8v0rmdM+JM+rNbaa4e4WRliiRW2l2YkLzggckdqu+JtXtoNKvLWa4Lebo yMvGTuMpUZP/AAL8hmvG7Bnur8JbEQW2neRIzbiNp4HJHP3j2r61QXsoNb2PJu+aSPrj4YzQWfjO ZYbqTbE6xpvYMzSblUgkDH94/hX0X8YdZu9JcvZyRJsYbflyM+/51+Y+m+Jb/RfHF7daRMojivGa Pem8N82R0xxivprQ/FOtePb+SfXWhNsGBHloVBb2yT/kiuOpgZz9/oHt1F2Y9fFd3rEvmSlg2Sz4 +XJq9d+J3tk2o27YuMg/dPvWwdGxcbYy6bRuyF5NJcusaMrmPcVwSy9f0rZYFfzAsTd7H5q38csV 2Yci5EWMvES45HYjNVp/D2pRzRSf2fdlmJkbKEblBHOOv8WK+zrjwfc3l+LnVdzTDChmk3HH1/Gt j/hGHsJJBPCGbkh87s9/6mv0x4qOzZ8T7N7nxBDo2oXVxAt7Y3eZZSpcruyeCefoRmvXLPTNSsPD sFpYaTfy3iymRpfsu5ccDbn2P4da+rND0G1kQNqqLFEcckZ5r6k8F+G7DUbVGtn86JSE+71HtXLL GqnqgdHmPyjb4e+J9Yw66bfyzEjYJCsYA65wa9H0D4D+Mr0H+z9Gj8+RSrSveMTz7Bfx4Nfrv/ZN uhjtbC3jDS4BUhcjHY98VZvZLPRLhhqzrF5aAnAKjPPHpXkyzWX2Ub/Vl1Pyxuf2PvF/iBXuNafQ rRlXawiWViR7jfgmuYf9mG6knFte69bqy/eVNPQjP/As+lfrMfiHpEWkRC4jRXuAdoZ8Ltzjn1PB r5a+J2tWs0DNoNkyT7icoc