MS TCP from Linux??

[Matthew Benjamin <MBenjamin@comshare.com>]

No offense, Brett, but Microsoft's TCP/IP implementation surely did not come
from Linux.  I am pretty sure Nicholas Pretreley has published that it,
along with internet utilities like ftp.exe, are ports from BSD 4.3.

Yes, that's right:  BSD.  But, that's not surprising, it was the reference
implementation for just about all of TCP/IP.  Microsoft had a variety of
choices.  Microsoft has owned the source code to at least one TCP/IP
implementation (Xenix--STREAMS?), and could have bought others at a bargain
price at any time.  Why use Linux's?  The Linux TCP stack of the day was not
competitive with a BSD stack, as Brett, the BSD advocate, must know.

The notion of clean room reimplementation is a plain red-herring.  Microsoft
need not clean-room re-implement any GPL'd code, or at any rate hasn't up
till now.

I can easily imagine a day when Microsoft might want to interoperate with
GPL'd code, maybe the CODA cacheing filesystem is an example.  (I choose
this example because such code (or its design) defines a communication
protocol, and to that extent resembles Brett's TCP example.)

However, this hypothetical, I think, actually argues in favor of the GPL:  

First, and quite typical these days, the CODA developers are doing ther OWN
Win95/NT port, so Microsoft would not need to do anything in order to
interoperate--unless it is the case that the NT filesystem driver APIs are
known within Microsoft to be suboptimal, in which case Microsoft could port
the GPL'd code itself.  The only drawback to this would be a reduction in
Microsoft's ability to obscure the structure of NT's driver subsystem, which
would be of benefit to consumers, in my view.  

Contrariwise, Microsoft's reason for making a proprietary implementation of
something like CODA would be, quite frankly, to change the protocol
specification, making it proprietary and maybe not interoperable with the
original.  (As they did with DCE RPC, Kerberos, etc.)  The new, "Microsoft
Cacheing Filesystem (MCF)" would give you All The Benefits of that Coda
filesystem, but with New, Enhanced Functionality, and Better Integration
with Windows NT!  (And would provide no or broken ability to share files
with non-Windows computer systems.)

This is exactly the sort of trick that the GPL prevents.  Yet, pace Brett,
this does NOTHING to restrict Microsoft's freedom to create new protocols,
formats, techniques, and code, all proprietary, until the end of time.  It
just prevents them from using GPL'd code as a starting point, and frankly, I
fail to see what is wrong with that.

I'll submit another opinion.  The sort of license that you prefer is
harmless when it is used by some network appliance vendor to drive their
canned system.  Who cares where Cisco got their source code?  (BSD, I've
heard, and more power to them.)  But in the UNIX system in the 1980s, the
use corporations put to their licensed source code--making every UNIX a bit
different from the others, while claiming compatibility, and denying users
access to the tools needed to fix problems not being fixed by the
penny-pinching vendors--had disastrous effects.  

This is what killed the original UNIX system--it is the "fragmented UNIX
market" that Microsoft has marketed against so effectively.  

Linux and other GPL'd systems are categorically less vulnerable to this sort
of fragmentation and stagnation.  That is a fact.  The only losers--and
there obviously are some, although the market was rapidly drying up when
Linux arrived on the scene--are those who want to be vendors of proprietary
UNIX systems (and Microsoft, which was transforming the decline of UNIX into
growth in Windows NT installations; since commercial Intel UNIX, including
(esp!) Microsoft Xenix, was so weak, yet just as closed and proprietary as
Windows, transitioning customers to NT was not difficult).  That is not a
loss for consumers, though, because it was the weaknesses (real and
perceived) of proprietary UNIX that were killing the market.  

Even then, I predict that although some UNIX vendors (Sun?) were committed
to managing the UNIX product top to bottom, as long as it sold Sparc
hardware, others (HP? Digital?) were under pressure to short-change UNIX
development, damaging product quality.  For these companies, Linux could
actually prove a better deal than their own proprietary blend:  they can
continue to sell value-added features (optimizing compilers, clustering,
enhanced filesystems, transaction control, etc) as applications or LKMs
running on the Linux kernel--while slashing kernel development overhead.  (I
leave out SCO, which in recent years has managed to be committed to total
proprietary control, and crummy implementation, and outrageous pricing, all
at once.  These faults, not Linux, constitute the threat to SCO's business.)

Now, it is indeed possible that, insufficiently rewarded for their creative
labour, the best Linux developers will just drop out, and work for industry
instead.  But until that time--and I see no sign of it happening yet--your
grand theories about creativity and the evil GPL are misplaced.  (In fact, I
predict the number of core Linux developers will increase, as commercial
Linux companies divert resources to making the Linux kernel suitable for
hosting their commercial features, to the chagrin of RMS, of course.)

History will decide this one, not rhetoric.


Matt

-----Original Message-----
From: Brett Glass [mailto:brett@lariat.org]
Sent: Saturday, January 02, 1999 7:06 PM
To: Steve Cohen
Cc: Multiple recipients of list AM-INFO
Subject: Re: Not "Satanism;" realism.


At 12:51 PM 1/2/99 -0600, Steve Cohen wrote:
 
>Well, I can see your point, but it does seem a bit farfetched to me.  I
can't
>think of a single piece of currently GPL'ed code right now that Microsoft
would
>want to make the effort to make a clean-room copy of, since, in many cases 
>these are reinventions of things Microsoft and other commercial developers
are 
>already selling or aren't interested in selling. 

Microsoft has already incorporated programs derived from open source
software
into its operating systems -- both Windows and NT. How much of Microsoft's
TCP/IP stack was clean-roomed from Linux? We will probably never know, but
the
fact that NT has been susceptible to many of the same DoS attacks
that plagued old versions of Linux raises some suspicions, at least in my 
mind.

>Of course that wouldn't apply to some new "killer app" that someone would
>hypothetically invent now and put under the GPL.  My sense though, is that
>corporations who are writing these kinds of apps now are not putting them
under
>the GPL.  With a few exceptions like Netscape (or Mozilla which still isn't

>GPL) many of them remain closed-source.  Let's not confuse the stated 
>intentions of Richard Stallman with what is actually happening.

Many of the stated intentions of Richard Stallman *will* actually happen if
we embrace the GPL.

--Brett