See also: last week's Back page page.

Linux links of the week

DocBook.org is the home site of the O'Reilly book DocBook: The Definitive Guide. The entire book is available from the site, as are errata, a DocBook FAQ, and more. Just about everything you need to get started with this growing standard for technical documentation is there.

Section Editor: Jon Corbet

Letters to the editor

From: "Allan Pointon" <allan.pointon@virgin.net>
To: <letters@lwn.net>
Subject: raccoon roti
Date: Fri, 3 Dec 1999 23:28:08 -0000

Q : what was the last thing that went through a certain recently deceased
raccoon's mind ?
A : Hummmm.........

Date: Mon, 06 Dec 1999 23:20:02 -0800
From: Pascal Martin <pascal.martin@iname.com>
To: letters@lwn.net
Subject: Linux demo at Microsoft's Comdex booth ?

Hello.

I wanted to share with you some funny news I just discovered.

Axis (http://www.axis.com/) is a company provided innovative
Internet Appliances, including an exciting Internet camera.

.. so exciting indead that they claim Microsoft made a demo
at their Comdex booth:

   http://www.axis.com/news/us/comdex.htm

The funny thing is that Axis also claims that this Camera
is based on Linux:

   http://www.axis.com/news/us/2100.htm

So, here we are, with some possibility that Microsoft might
_really_ have done a Linux demo at Comdex !!!. And we get
the ultimate pleasure of seing together on the same web
pages these two related news (enjoy !):

   http://www.axis.com/news/

What do ya think ?  :-D

------------------------------------------------------------------
Pascal F. Martin.

Date: Wed, 08 Dec 1999 06:29:42 -0900
From: "Tony Taylor (ISD)" <tony@searhc.org>
Subject: LinuxToday Zope Review
To: kreichard@internet.com, letters@lwn.net

Mr. Reichard,

In your review of the Zope application server, you raised many valid and
important points.  However, one criticism was completely unfair and
irrelevent.  The strawman you erect is in this paragraph:

"In addition, the application-server market has largely settled on Java
and Enterprise JavaBeans (EJBs) as tools that can be deployed on any
application server, allowing for different servers from different
vendors to interoperate on a useful level. There's no hint at all within
Zope or the Zope documentation that Java/EJB support was ever
considered, much less implemented. This is a major drawback within Zope;
you can argue all day and night that Python is superior to Java, but one
simply can't ignore the realities of the marketplace."

Only two years ago, the marketplace insisted MS-Windows NT was the
dominant applications platform.  In spite of this market reality, Linux
has grown in popularity not by embracing NT, but by its simple technical
superiority.  I will not argue the superiority of one language over
another; I personally dislike much of the Python syntax.  But to declare
a product unfit simply because it does not follow popular market trends
is myopic and unimaginative.  Java is not a standard, and is apparently
never going to be a standard.

Some of the other criticisms are spot-on; I only disagree with arguments
like, "A is popular; B doesn't support A; therefore, B is bad."  If
popularity were any measure, Budwieser would be a good beer, and NT
would be a decent operating system.

					- Tony

Date: Wed, 08 Dec 1999 01:06:17 -0500
From: Derek Glidden <dglidden@illusionary.com>
To: j2se-linux-comments@sun.com, letters@lwn.net
Subject: Sun, Java and Linux


Today on the Java-linux mailing list there started a flame-fest because
today Sun announced that they, jointly with Inprise (nee Borland), have
released the Java2 JDK v1.2.2 Release Candidate 1 for Linux,
coincedentally on the same day the Blackdown team announced their Java2
JDK v1.2.2rc3.  The flame-fest comes from the fact that Sun's press
release conveniently doesn't mention the Blackdown group who have been
hacking away at the Java2 source code since pre-release versions to get
it running successfully on Linux.  

I follow java-linux development more than I follow linux-kernel
development, but from the way I understand things, a good deal of
native/kernel thread development and stability on Linux is due to the
Blackdown team working with the kernel and glibc teams to find and track
down threading bugs in the kernel and glibc.  Without the work Blackdown
has done, a full Java2 JDK on Linux that could take advantage of native
kernel threads just might not have been possible at all.  Without
Blackdown pushing, it's also possible that Linux kernel hackers might
not have had impetus to implement native kernel threads at all by now. 
Linux has more to thank the Blackdown team for than just the JDK port.

Adding insult to injury, when you download the "Sun/Inprise" JDK from
Sun, you find READMEs and CVS messages left in several of the files left
over from the Blackdown port. In other words, Sun took the work the
Blackdown team has been doing on Java2 for *at least* the past year,
handed it to Inprise, said "Here, get this running" and turned around
and claimed it as their own.  In their defense, Inprise says they've
done a lot of independent work on the JDK to get it to "release quality"
but it's still founded on man-years of work done by Blackdown.  (And the
Inprise port doesn't support native threads and actively discourages
running it on SMP boxen, while the Blackdown port does both just fine,
so...)

Unfortunately, Sun's "Community Source License" gives them every right
to do exactly that and screw Blackdown and screw their porting efforts
in the process.  The Blackdown team has been working on the JDK in good
faith, and against nearly overwhelming apathy from Sun, knowing full and
well that Sun's source license gives Sun the exclusive right to any and
all patches to the JDK source that Blackdown might generate, under the
assumption that the Blackdown port would be Sun's "semi-official Linux
JDK" port. For their effort, Sun turned around and said, "Screw good
faith and screw you too, we're gonna get some good press out of this and
pretend again like we're an 'Open Source' company and 'Pro-Linux' just
like when we bought StarOffice.  We're Sun, you're a bunch of nameless,
faceless programmers.  Who's spin do you think the public will
believe?"  

Not only that, but the press release goes so far as to say (paraphrased)
"Java and Linux fans have asked for Sun to support Java on Linux and we
listened".  This after SEVERAL YEARS (ever since JDK 1.0) of "native
Linux support from Sun" being the most demanded feature on Sun's "JDK
wishlist" by SEVERAL ORDERS OF MAGNITUDE over ALL OTHER DEVELOPER
REQUESTS COMBINED and Linux users getting nothing but the finger from
Sun.

Sun - Your license stinks.  Your concept of "community" stinks.  The way
you treat (use) developers trying to support your products stinks. 
Forget your mouth, this time you've stuck your foot all the way up your
a** after having shot it nearly clean off and there's no one to blame
but yourselves.  Hopefully the tone of this letter will give you some
understanding of the bad feelings ("bad feelings" isn't strong enough,
but I'm trying to remain more or less polite) you've generated in the
Java-Linux community.  Good luck trying to regain our trust.  

On the very dim bright side, a member of the Java development team from
Inprise is also (unofficially) frequenting the java-linux list and is
(again, unofficially) trying to let the world (or at least the
java-linux world) know that, at the very least, Inprise developers know
full and well the effort Blackdown has already put into the work and
that Blackdown has Inprise's full and complete respect for it. 

That and a buck and a half will get the Blackdown team a cup of java
down at Starbucks and big fat moon from Sun.
 
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
With Microsoft products, failure is not           Derek Glidden
an option - it's a standard component.      http://3dlinux.org/
Choose your life.  Choose your            http://www.tbcpc.org/
future.  Choose Linux.              http://www.illusionary.com/

Date: Wed, 8 Dec 1999 23:09:23 +0000
From: Alain Williams <addw@phcomp.co.uk>
To: lwn@lwn.net
Subject: Comment on No FUD suggestion by Linuxtoday

First let me say that I am supportive of what they are doing.

I have a suggestion that, at first, may seem strange and counter to what
you are doing but which, in the long term, may make the arguments much more
powerful.

What they are setting out to do is to ``dispell the shadows of doubt with the
clear light of truth''. If this isn't exactly what they are about then maybe
this is what they should do. What is the difference ? *Honesty*. If we are
completely honest no one can trip us up.

How will honesty change what they are doing ?
Basically it means that faults/problems/... with Linux have to be openly
pointed out. If faults are pointed out it makes the good points talked
about have much more credibility. This is the antithesis of much marketing
where a veil is drawn over deficiencies in whatever they are talking about,
techies/hackers distrust marketing people for exactly this reason: techies
are (by & large) a pretty honest bunch.

So what do I think NoFUD should be doing ?
Also have anti FUD pages about other operating systems: anything from Big
Blue's VM to Microsoft's offerings.

Why will that help ?
1	It will improve the credibility of the NoFUD web site.
2	It elevates NoFUD from a ``Linux Site'' to an ``Industry Reference
	Site''.
3	It will be visited by many more people, people who are after
	genuine insight into advantages/disadvantages of the various
	platforms that are discussed.

How will this help Linux ?
It will show people where Linux performs well (or badly) in relation to
the other OSs that are out there. The places where Linux can do with
improvement will he high-lightened and (hopefully) attract more effort from
the Open Source community.

-- 
Alain Williams

Date: 3 Dec 1999 01:30:26 -0000
From: mogul-linux@gelatinous.com
To: letters@lwn.net, malda@slashdot.org
Cc: products@opengl.org
Subject: Fahrenheit, SGI, Microsoft, Linux, D3D, OpenGL


I'm an experienced game developer.  I read this story about Microsoft
quietly dropping OpenGL support in Windows 2000:

    http://www.theregister.co.uk/991129-000007.html 

It also details the failed alliance intended to draw Direct3D and OpenGL
together... After reading it I was inspired (incensed?) enough to write
down the events in this sordid affair as I remember them.  The Linux
reference near the end of the article makes the timing of this move
obvious, but I've always felt that Fahrenheit was smoke and mirrors to keep
game developers away from vendor-independent OpenGL and using a standard
Microsoft could control.  See if you feel the same way after you see my
(admittedly biased) list of the facts in order:

    Despite WinNT supporting OpenGL happily from near its inception,
	 Microsoft wastes the first three versions of DirectX trying to
	 come up with a usable 3D API for Win9x, claiming the OpenGL API is
	 not suitable for game development.

    Apple announces that they will provide Game Sprockets for the Mac, a
	 suite of game development libs similar to DirectX.  Rave3D is
	 incorporated as the 3D API.

    Microsoft claims that they will be supporting DirectX across every
	 operating system they can, including their own, MacOS, HP-UX,
	 Irix, Solaris, etc.  They say they expect DirectX to become a
	 standard, and that developers shouldn't worry about compatibility
	 because wherever they want to take their games, DirectX will be
	 supported. Therefore, there's no point in programming to Rave3D.

    OpenGL vs. Direct3D debates rage across the industry for months, with
	 many of the harsher criticisms of D3D centering on the fact that
	 the API makes optimization of drivers nearly impossible.  D3D
	 provides only capability bits to test for features, and guarantees
	 none.  OpenGL is criticized for being slow, large, and unsuited to
	 consumer-level cards, as well as inflexible due to the pool of
	 vendors who develop it by committee.

    Microsoft says it will stop debate by supporting both standards.  Soon
	 thereafter it supplies its implementation of OpenGL for Windows
	 9x, a ridiculously slow and incomplete version that breaks the
	 number one promise of OpenGL: all features are available, even if
	 they are only implemented in software.  It's even slower than
	 their software-only D3D implementation.  Supporting this version
	 of OpenGL is more difficult than supporting software D3D since
	 there aren't even capability tests to provide workarounds.  Many
	 developers throw up their hands and go for D3D since their games
	 have to ship someday.

    Around this time, Quake goes hardware accelerated with an OpenGL
	 wrapper on the 3dfx chipset, utilizing 3dfx's glide library.  John
	 Carmack's considerable influence over the industry and strong
	 cross-platform stance makes waves, and with GLQuake as the poster
	 child, the debates start having conclusions: OpenGL *can* be
	 lightweight enough for game development, and provides much more
	 opportunity for optimization to card vendors.

    SGI steps into the fray by offering a free fully-compliant and even
	 well-optimized software OpenGL replacement for Windows 95, with an
	 attendant architecture that makes it possible to write installable
	 client drivers for hardware acceleration fairly easily.

    Apple, losing money fast and on the brink of death, seeing the already
	 dwindling Mac game market shrink even further, stops supporting
	 Game Sprockets and redirects the Sprockets developers to their OS.
	 They go silent for a while.  However, 3dfx ships a PCI version
	 that supports glide on the Mac.  GLQuake is ported to the Mac
	 using the 3dfx OpenGL wrapper, and runs great!

    Microsoft makes dramatic "updates" in the shaky D3D API.  It drops its
	 poorly conceived and executed retained mode (a very poor imitation
	 of SGI's Performer functionality), and focuses exclusively on
	 duplicating the functionality that OpenGL has had all along
	 (discouraging raw vertex buffers, adding DrawPrimitive, etc.)
	 This isn't enough for developers, who have all that plus the added
	 bonus of portability when they use OpenGL.  Microsoft and SGI
	 start making lots of noise and lawsuits start flying.

    Coincidentally, many companies (such as Intergraph) have been steadily
	 eating away at SGI's hi-end market share over the last few years
	 as they make OpenGL workstations running OpenGL on Windows NT that
	 rival mid-range SGIs.  In a sudden turnabout, SGI announces that
	 they'll turn their OpenGL and ICD implementation over to Microsoft
	 to incorporate and support in place of Microsoft's shoddy version.
	 The two companies announce that they'll work on a next-generation
	 hybrid API called Fahrenheit that incorporates the best features
	 of D3D and OpenGL and adds Performer functionality.  Note that
	 ONLY Microsoft and SGI are involved... The million other companies
	 in the OpenGL Architecture Review Board are left out in the cold.
	 Shortly thereafter, SGI announces that it will be making WinNT
	 workstations as well.

    Peace is declared for a while as developers sit through DX5, DX6, and
	 DX7, waiting for an announcement any day that the first version of
	 Fahrenheit is ready for testing.  Developers prefer OpenGL due to
	 the caps bit issue, but ship versions that run with both OpenGL
	 and D3D so that they're supported by as many cards as
	 possible. (Many don't even support D3D until the game is done,
	 then spend the rest of their time on compatibility issues in their
	 D3D support.)

    Meanwhile, the OpenGL ARB starts moving much faster with the addition
	 of companies such as NVidia and 3dfx whose core hardware engineers
	 used to work at SGI, and whose excellent products are dominating
	 the consumer market.  Extensions to the standard start churning
	 out at a pretty good clip, keeping pace with new features coming
	 out in consumer graphics cards such as multitexturing.  Quake
	 becomes something of a performance benchmark for these leading
	 consumer cards, and companies such as ATI suddenly have to support
	 OpenGL in order to be able to compare apples to apples.  Suddenly
	 every card vendor for the PC is supporting OpenGL.

    Apple announces that the iMac will incorporate ATI's 3D cards and
	 support OpenGL.  The new machines are zippy and suddenly the Mac
	 looks like a great game platform, especially if you're doing
	 OpenGL development on the PC anyway. For these amid other reasons,
	 Apple is back on top of its game again.

    Microsoft encourages hardware vendors not to ship OpenGL installable
	 client drivers, saying that their effort would be better spent
	 writing drivers for the Windows Driver Model architecture, the
	 unified driver standard for Windows NT and Windows 2000.
	 Companies would of course love to support one driver only, so
	 their ICDs are shipped but without much optimization.  Many say
	 "wait for Win2K, where our driver support will be great!"
	 (eg. Matrox)

    Mesa (a free software OpenGL implementation) and Linux have been
	 behaving well together for a while, with some hardware
	 acceleration based on a Linux version of the glide library from
	 3dfx.  Linux makes a couple of usability leaps that give Microsoft
	 pause. In early 1999, NVidia and 3DFX announce custom binary-only
	 X servers that support their cards.  SGI suddenly announces
	 massive support for Linux, and open sources it's GLX library for
	 the purpose of incorporating Mesa into the free X server,
	 XFree86. NVidia, now the market leader, says it will have fully
	 accelerated OpenGL support for all of its cards under Linux by the
	 end of 1999.

    Microsoft finally tries ditching OpenGL altogether in Windows 2000, as
	 detailed by the article.


We all know how roughly Microsoft plays in the industry, but watching this
progression has really upset me.  OpenGL has always been a cross-platform
standard carefully supported and developed by multiple vendors.  The
various attempts on Microsoft's part to wrest control are pretty obvious:

	Bring out a competing standard, though inferior to the existing one.
	Claim to be providing a standard for the entire industry, not just
	    themselves. 
	Bad mouth the capabilities of the competing standard.
	Claim to support both, but provide crappy support for OpenGL.
	Desperately try to incorporate the features they lack into their own
	    standard. 
	Bully beleagured SGI into giving them control over the better version.
	Get driver writers dependent on their ICD arch, then sandbag driver
	    development. 
	Drop the Fahrenheit architecture carrot that got everyone moving their
	    way in the first place.
	Quietly drop support for the competing standard at the last second,
	    just when it's again obvious why OpenGL is the industry standard.

No one should be surprised at this maneuver after all of Microsoft's other
activities, but it makes me wonder why ANY company is still willing to work
with Microsoft after seeing the way they attempt to eat not just their
competitors, but their allies as well.  Is it any wonder hardware vendors
are fleeing to Linux as sanctuary from the ravening beast?

mogul-opengl@gelatinous.com