01.12.09
Bill Gates: “Where Are We on This Jihad?” (Against Linux at Intel)
In previous posts we mentioned Intel because Microsoft was trying to turn Intel’s attention to Windows, at the expense of Linux. The following exhibits are what one reader called “the deep concern at Microsoft when Intel switched their internal industrial dev OS to GNU/Linux.”
Added at the bottom we have exhibit px07022 [PDF] and exhibit px07068 [PDF] from Comes vs Microsoft, under Appendix A and Appendix B, respectively.
With help from an unnamed reader we highlight bits of the exhibits that show the relevance of this to GNU/Linux.
Here is an E-mail Bill Gates wrote to Eric Rudder, whose take on .NET we've also just mentioned.
—- Original Message —-
From: Bill Gates
Sent: Wednesday, November 21, 2001 12:07 PM
To: Hike Porter; Eric Rudder
Cc: Jim Allchin; Steve Ballmer; David W. Williams; Brian Schuster
Subject: RE: My personal take on INTELI went and read the enclosure
I think we will have to live without a Chinese wall clause for the front end of the compiler.
It’s too bad that all source code isn’t Chinese wall but since Intel views their processor specifications as equivalent to source code we don’t want all of these things subject to Chinese wall restrictions.
If we can move ahead having the CLR and backends having Chinese wall provisions then we should take the risk relative to the front end.
They can’t still move our trade secrets or code to another front end but they can have common people work on those things.
Are there other issues outstanding here? Did I misunderstand the discussion?
One reader, who calls this “the Intel Chinese non-Wall,” explains that what Gates essentially tells is: “Let’s wait until we get a look at the processor specs before invoking a Chinese wall on the compiler. If Intel raises objections we’ll say the non-Chinese proviso only applied to the front-end.”
Further, we have:
From: Eric Rudder
Sent: Tuesday, November 20, 2001 8:08 AM
To: Bill Gates; Steve Ballmer
Cc: Jim AIIchin; Christine Turner; Dorothy Veith; Debbie Hill
Subject: My personal take on INTEL
Importance: High[...]
We do not have a signed agreement yet with Intel to work on the CLR and Compiler. I am actually optimistic we will close on this, but it’s ridiculous that it languished for 10 months before I got here, and I have had to personally spend hours driving.
We need help on:
- Floating Point
- CLR
- 64-bitFP is a case where we have fallen behind Linux, thanks to Intel’s great work w. Linux compilers. They can help us with the Math libraries and some OpenMP stuff. We want access to their benchmark/test suites. it’s crazy that we can’t get Intel to do Windows first, then Linux (if they must)
[...]
2. If we don’t get Intel off of Linux internally (the failed EDA project) – we will never get the *cultural* alignment that we want. There are simply too many folks at Intel who use/love the stuff and want to improve it. We can *not* stop trying to win this project.
A reader’s interpretation is as follows: “Let’s get Intel to clone their Floating Point Linux ‘stuff’ on Windows, as a stratagem to kill internal Intel Linux projects, let’s get Intel to transfer their people to our project.”
Also interesting is the following part:
I also have a personal ask that if we do the 64bit work that we get INTEL.COM to be a 64bit ASP.NET sites.
Intel runs GNU/Linux in its Web sites, so this favour never worked out on the face of it.
The second exhibit about this is concerned with “getting Intel to dogfood Windows,” to use Microsoft’s own words. Bill Gates inquires, “Where are we on this Jihad?” Yes, Microsoft has used the term "Jihad" -- albeit internally -- on quite a few occasions. Gates also can’t spell Otellini’s name (Intel’s CEO). He writes “Ottelini”.
Here are some highlights that refer to content rather than vocabulary:
From: Brian Valentine
Sent: Saturday, February 02, 2002 5:23 PM
To: Srini Koppolu; S. Somasegar, Bill Veghte; Vinod Anantharaman, Doug Miller, Chris Ray; Eric Rudder; Ann Marie McLeod
Cc: Jim Allchin
Subject: RE: Intel EDA migrationPlease make sure Ann Marie is on these threads – I have added her here.
As far as going after them — they are important from the perspective of getting Intel to dogfood Windows. This would be a big thing we could both talk about, etc We want to them on Windows We do need to look at the all the ISVs, etc and make sure that we have good programs in place to move them.
[...]
—– Original Message —-
From: S. Somasegar
Sent: Friday, February 01, 2002 4:35 PM
To: Bill Veghte; Vinod Anantharaman; Doug Miller; Chris Ray; Srini Koppolu; Brian Valentine; Eric Rudder
Subject: RE: Intel EDA migrationI agree on ROI, etc.
With Intel though, it is a crime that these guys are running Linux and it is a shame that we can’t get them to move to Windows for their core development systems. I also think that unless it is a top-down initiative at Intel to “just make this happen”, this will not get traction no matter how much we try.
I happen to know that Intel engineers are indeed using GNU/Linux (in part) because my brother in law works there. Overall, this second exhibit makes Microsoft look pitiful. It’s chasing Intel and the EDA ISVs just don’t want Windows. The same goes for Intel, which talks about disdain for Windows NT (“EDA ISVs got burnt with poor experiences with Windows NT, are wary of taking steps in this direction”). █
Appendix A: Comes vs. Microsoft – exhibit px07022, as text
From: Mike Porter
Sent: Wednesday, November 21, 2001 1:47 PM
To: Peter Biddle
Subject: FW: My personal take on INTEL
Tracking: Recipient Read
Peter Biddle Read: 11/21/2001 1:58 PM
It’s been a long week
—- Original Message —-
From: Bill Gates
Sent: Wednesday, November 21, 2001 12:07 PM
To: Hike Porter; Eric Rudder
Cc: Jim Allchin; Steve Ballmer; David W. Williams; Brian Schuster
Subject: RE: My personal take on INTEL
I went and read the enclosure
I think we will have to live without a Chinese wall clause for the front end of the compiler.
It’s too bad that all source code isn’t Chinese wall but since Intel views their processor specifications as equivalent to source code we don’t want all of these things subject to Chinese wall restrictions.
If we can move ahead having the CLR and backends having Chinese wall provisions then we should take the risk relative to the front end.
They can’t still move our trade secrets or code to another front end but they can have common people work on those things.
Are there other issues outstanding here? Did I misunderstand the discussion?
—– Original Message —-
From: Mike Porter
Sent: Wednesday, November 21, 2001 11:47 AM
To: Bill Gates; Eric Rudder
Cc: Jim AIIchin; Steve Ballmer; David W. Williams; Brian Schuster
Subject: RE: My personal take on INTEL
Additional hardware is already on the way, but I think they are all ear-marked for the NT Base team. I will research. Eric, we can get you systems. I’ll ask for 200, but we’ll see what we can get… production is much more Iimited on McKinley than on Merced. They are severely constrained… we have the vast majority of what they can produce onsite.
Eric, I think you are dead on about the EDA project. BrianV drove this hard, had an entire team in place to address every issue they came up with… and they did completely address that list: Intel still didn’t migrate. I believe that was at least an 18 month project led by Stephanie Boesch, who did a great job. Intel did seem to express renewed interest in pushing their internal org here in the exec review yesterday. It would be a huge step in fixing the cultural issues. That’s why we’ve been pushing them for over 2 years on this at Bill and Steve’s level.
MS-CC-MDL 000000480159
HIGHLY CONFIDENTIAL
Eric, we should set up a meeting between you and Renee, who control allocation of the resources (bodies and systems). She’s asked to meet you and she needs to hear your concerns. Well try and set that up in the next week.
The patent license will help, but it will not eliminate many of the specific concerns that are coming up, and won’t solve any of Eric’s stated concerns. We received our first draft of the compiler DA on June 8th. To my calculations, that means 5 months. I’m not sure where the 10 month figure came from. We thought we were ready to sign this week, with an agreement based on our standardized template agreement, but Eric asked us to go back to Intel and re-negotiate one more time for an additional change to the Chinese firewall provision. I pushed back this morning in email to Eric that we are at a point were doing what Eric suggests will set a precedent we have jointly agreed not to set with Intel at a broader relationship. We don’t ask for Chinese walls on anything except “Crown Jewels” (i.e. compiler backends, OS kernels, Intel RTL “silicon” code, etc.) or Intel would reciprocate (based on discussions with them) and try to put the same Chinese firewalls in place with us on their led book materials (and they’d LOVE to do that to us given our work with AMD) I told Eric I’d go back to Intel, bet only if JimAII or BilIG agrees that this is worth risking in the overall context of the broader relationship, knowing full well that we will get asked to start doing the same thing and they will have to get involved in the future. Details on the specifics of Eric’s issues are in the attached email thread. I am still waiting far a decision on how to proceed here.. which is only adding time to clock.
—– Original Message —–
From: Bill Gates
Sent: Wednesday, November 21, 2001 11:07 AM
To: Eric Rudder
Cc: Jim AIIchin; Mike Porter; Steve Ballmer
Subject: RE: My personal take on INTEL
This email was very helpful.
We expect to close some kind of patent cross license in the next month which should help get things going better.
We used the example of helping us with our compilers as a prime example where things should be easy to get going.
Not getting hardware surprises me. Mike Porter should be able to help with that.
— Original Message —
From: Eric Rudder
Sent: Tuesday, November 20, 2001 8:08 AM
To: Bill Gates; Steve Ballmer
Cc: Jim AIIchin; Christine Turner; Dorothy Veith; Debbie Hill
Subject: My personal take on INTEL
Importance: High
I know you are meeting w. Intel today – here are some thoughts I would like to share and some status. (Chris, D/D, if Bill and Steve can see this mail before the mtg, that would be great.)
1. It’s way too hard to get anything done w. these guys. We need a new master agreement, where a PM can simply work off a template to create a new “project.” in the absence of this it takes MONTHS to get things done that should literally take minutes.
We do not have a signed agreement yet with Intel to work on the CLR and Compiler. I am
2/21/2003
MS-CC-MDL 000000480160
HIGHLY CONFIDENTIAL
actually optimistic we will close on this, but it’s ridiculous that it languished for 10 months before I got here, and I have had to personally spend hours driving.
We need help on:
- Floating Point
- CLR
- 64-bit
FP is a case where we have fallen behind Linux, thanks to Intel’s great work w. Linux compilers. They can help us with the Math libraries and some OpenMP stuff. We want access to their benchmark/test suites. it’s crazy that we can’t get Intel to do Windows first, then Linux (if they must.)
The CLR work would mostly be around optimization for XScale and tuning. Intel may say they are doing work around this – the truth is we are seeing very little results.
64bit is the biggest no-brainer. In addition to help from people (for both compiler and JIT), we also have an ask for hardware. My group wanted me to ask for 200 boxes! I think this is too many, but even 25 could make a big difference (and yes, we are smart abt using terminal server, etc — we have many, many configs.) I also have a personal ask that if we do the 64bit work that we get INTEL.COM to be a 64bit ASP.NET sites.
2. If we don’t get Intel off of Linux internally (the failed EDA project) – we will never get the *cultural* alignment that we want. There are simply too many folks at Intel who use/love the stuff and want to improve it. We can *not* stop trying to win this project.
My two cents.
-Eric
2/21/2003
MS-CC-MDL 000000480161
HIGHLY CONFIDENTIAL
Appendix B: Comes vs. Microsoft – exhibit px07068, as text
From: Brian Valentine
Sent: Saturday, February 02, 2002 11:45 PM
To: Mike Porter; Bill Gates
Cc: Jim AIIchin; Ann Made McLeod
Subject: RE: Intel’s design engineers and Windows
We are all over them on this and will keep trying to get them to move. Attached is the whole thread that is currently running on this.
—– Original Message —–
From: Mike Porter
Sent: Thursday, January 31, 2002 12:34 PM
To: Bill Gates
Cc: Jim AIIchin; Brian Valentine
Subject: RE: Intel’s design engineers and Windows
Sorry It’s taken so much time to investigate. BrianV put a team together to nail this and in our opinion we did everything possible and had a pretty solid solution. Intel said they felt we didn’t meet every need, although I’ve been pushing for the list of “how we failed them” for over 2 weeks now. Bottom line, IMHO, Intel doesn’t want to deal with their internal politics and “sell” this transition internally. Think about our development org. bright, extremely talented and opinionated folks… and at Intel, the developer crowd was raised on Unix/Linux. They just don’t WANT to move. Also, this was originally being driven by Albert, and that changed to Gelsinger for a while I’m not sure given their recent changes internally who owns this. This may be an area that Otellini could help.
Given that you are meeting with Paul on the 14th of this month (and we have a prep meeting on the 6th), is there any data I can get you to aid in your discussion with Paul?
—– Original Message —–
From: Bill Gates
Sent: Monday, January 14, 2002 1:41 PM
To: Mike Porter
Cc: Jim AIIchin; Brian Valentine
Subject: Intel’s design engineers and Windows
Where are we on this Jihad?
Do I need to be calling and emailing Ottelini to get this back on track??
Every day that goes by is a bad one for us on this. Despite the difficulty we need to draw the line in the sand on this one for a lot of reasons.
MS-CC-MDL 000000406401
HIGHLY CONFIDENTIAL
From: Brian Valentine
Sent: Saturday, February 02, 2002 5:23 PM
To: Srini Koppolu; S. Somasegar, Bill Veghte; Vinod Anantharaman, Doug Miller, Chris Ray; Eric Rudder; Ann Marie McLeod
Cc: Jim Allchin
Subject: RE: Intel EDA migration
Please make sure Ann Marie is on these threads – I have added her here.
As far as going after them — they are important from the perspective of getting Intel to dogfood Windows. This would be a big thing we could both talk about, etc We want to them on Windows We do need to look at the all the ISVs, etc and make sure that we have good programs in place to move them.
—— Original Message ——
From: Srini Koppolu
Sent: SatIJrday, February 02, 2002 6:46 AM
To: S. Somasegar; Bill Veghte; Vinod Anantharaman; Doug Miller; Chris Ray; Brian Valentine; Eric Rudder
Subject: RE: Intel EDA migration
It would have been worse if we haven’t got the CAD ISV apps working on Windows. Atleast no one can make an argument that the Windows is not ready for high end apps
Let’s engage with Intel for some more time and do some research on EDA space. But if it is a no go, we better cut our losses soon and not get into rat hole with Intel for months/years like in previous cases.
—– Original Message —-
From: S. Somasegar
Sent: Friday, February 01, 2002 4:35 PM
To: Bill Veghte; Vinod Anantharaman; Doug Miller; Chris Ray; Srini Koppolu; Brian Valentine; Eric Rudder
Subject: RE: Intel EDA migration
I agree on ROI, etc.
With Intel though, it is a crime that these guys are running Linux and it is a shame that we can’t get them to move to Windows for their core development systems. I also think that unless it is a top-down initiative at Intel to “just make this happen”, this will not get traction no matter how much we try.
—- Original Message —-
From: Bill Veghte
Sent: Friday, February 01, 2002 4:31 PM
To: Vinod Anantharaman; S. Somasegar; Doug Miller; Chris Ray; Srini Koppolu; Brian Valentine; Eric Rudder
Subject: RE: Intel EDA migration
Of all the different Unix migration targets, I am pretty skeptical that EDA is the most leveraged for us to go after where
Leveraged = (we can win with reasonable investment)+ (large economic return for MS relative to investment)+ (big credibility boost).
2/5/2003
MS-CC-MDL 000000406402
HIGHLY CONFIDENTIAL
I am much, much concerned about all the different corporate LOB apps on wall street, insurance, etc. those are the targets where we want to win and get the PR around.
—– Original Message —-
From: Vinod Anantharaman
Sent: Thursday, January 31, 2002 5:49 PM
To: S. Somasegar; Doug Miller; Chris Ray; Srini Koppolu; Brian Valentine; Bill Veghte
Subject: Intel EDA migration
Summary of a conf call Doug & I had with Intel today, re: migrating their ECAD / EDA environment from Linux/UNlX to Windows.
Background
* Intel President Paul OtteIlini asked his team to work with MS, figure out what it took to move their EDA environment to Windows. Soma & Renee James at Intel had a recent discussion on this
* Intel ran a EDA migration project back in early 99, to move to NT4 – they got badly burned on this due to stability & software migration issues, and recommended going with Linux instead.
* MS got involved around mid 99, did a joint project where Intel evaluated Wfn2k + SFU vs. Linux, and a list of about 100 or so MS work-items were identified. In the end, after 18 months of engagement or so, this didn’t pan out- some of the reasons are listed in the first email attached above. Intel went ahead w/ the Linux route.
Attendees
* MS: DougMil, VlnodA
* Intel: Greg Spirakis (VP, ECAD Design Tools), Elwood Coslett & Kevin Wheeler (program managers in Engg. Compufing / IT), Mike Webb (don’t have has designation, he facilitated this conf)
Meeting Summary
Intel summarized their migration requirements thus:
o Primary order bit is that MS must convince the top EDA players to move over to Windows. Intel firmly believes this is what it takes for them to consider moving to Windows. They were very crisp/ up-front on this point.
o They use about 100 odd tools from 8-10 different vendors, will share the list with us (will include the usual cast of characters: Cadence, Synopsis, Unigraphics,..)
o Additionally, there’s a big ecosystem of tools & scripts they’ve developed around the major ISV apps, so all of these things have to be migrated as well. Once the ISV apps become available on Windows, they would need to find resources for the Windows migration, and as they said “take the risk associated with such a switch-over”
Intel’s perspechve on what’s changed since we last engaged with them:
o EDA ISVs got burnt with poor experiences with Windows NT, are wary of taking steps in this direction
o ISVs have been able to move to Linux easily – ported their code more easily, able to share code b/w UNIX / Linux, interop story is good. So they’ve been able to get cost benefits of IA hw w/Linux as a viable alternative OS
o Chicken & egg problem that ISVs still see no customer demand for Windows versions
o ISVs are trying to reduce the total # of platforms they support
2/5/2003
MS-CC-MDL 000000406403
HIGHLY CONFIDENTIAL
cuts their R&D, support etc. costs. Ideal scenario for them is that UNIX disappears and they support only Linux.
Intel’s perspective on what’s not changed since the last time
o Continued need for interop (b/w UNIX and Windows, in our case)
o Continued need for stability of environment, OS, shell environment, scripts, etc.
Linux apparently meets over 90% of their current EDA needs. They indicated a few aspects where Windows does better (but qualified these as being “less critical” requirements)
o Intel developers prefer using Windows / VS as their dev environment for all their ANSI C, C++ apps
o IT would prefer to support a single OS platform, and they clearly need their Windows desktops
o Better integration with Windows desktop / Office, although they said they are able to do this OK with their current X based solution
When asked to put themselves in our shoes, here’s what they said they’d do (nothing earth shattering here…)
o Find a unique value prop that will convince EDA ISVs about the advantage of supporting Windows & .NET. They said they’re happy to help us with develop this, since they’re familiar with the terrain.
o Point out MS successes in related spaces like mechanical CAD etc. & how we created value in those ecosystems
Next Steps
- We’ll get the list of key EDA ISVs that Intel depends on
- We’ll get feedback from Intel & their customers on Windows-UNIX interop issues, feature ideas, etc (some of their customers mentioned some issues here, note that Intel is currently on the SFU 3.0 beta program)
-In light of Intel’s position wrt to EDA ISVs, the right folks at MS (EPG?) should revisit the issue of whether we want to go after that business again Clearly this is a long term / uphill battle, if we want to go after it. I’ve attached a second mail thread where this topic was raised m March 2001, no decision was made back then to pursue this market aggressively. There was also an associated PPT Chris Ray and co. put together regarding the EDA space, I’m happy to fwd to anybody who’s interested.
-Vinod
2/5/2003
MS-CC-MDL 000000406404
HIGHLY CONFIDENTIAL
