pellucid, Nth Graphics, DuPont Pixel Systems, ATT/USL, and 9 more. SGI has
stated it intends to begin shipping OpenGL on its systems in Q1 1993. IBM and
DEC have stated their intention to release OpenGL in roughly the same time
frame. Nth Graphics will have OpenGL for HP and Sun workstations in early
1993. DuPont will have a Sun version of OpenGL in early 1993. Microsoft has
announced its intention to ship OpenGL as a standard part of Windows NT2, in
late 1993/early 1994. Since the OpenGL and GLX sample implementation and related
source code was released in June 92, by the end of 1993 we should see products
from many of these 22 companies.
A key distinguishing feature between PEX and OpenGL will be OpenGL's
availability in markets other than Unix/X Windows - i.e the PC and the Mac.
While many on the net seem to think the world begins and ends with X/Unix (hey,
lots of people here do!), ISV's want a standard that provides them access to as
many large markets as possible - and the PC and Mac are very large potential
markets for 3D technology. Remember, X Window based systems represent 3% of the
desktop computing marketplace. PC's and Mac's represent the rest. Windows NT will play a big role in the PC market. BYTE Magazine's cover (Is Unix Dead? - September 1992) suggests they feel NT will play a big role in the WS market as
well.
|>
|> A sample PEX server is included with X11R5. Most vendors now
|> have PEX server products for their workstations.
We are currently aware of 7 shipping commercial PEX products. These products
are based on the 5.1 PEX protocol, to be superceded by the PEX 6.0 protocol,
which becomes
available commercially according to best estimates in early 1994. While
we are aware of other companies intending to support PEX 5.1/6.0, their products
will likely be released in late 1992/early 1993 - not really much of a time
difference versus OpenGL when you consider the length of an ISV adoption cycle.
Many ISV's will be reluctant to develop code on 5.1 given the incompatible
protocol change coming up with 6.0. Many ISV's could choose to hold off on
PEX development until 6.0 is released and stable. An argument could be made
that the 'complete' version of OpenGL will be available prior to the 'complete'
version of PEX.
|> PEX Terminals (X-Terminals that support PEX) are on the market. This means
|> that PEX clients are assumed of widely available, relatively
|> vendor-independant and cheap (PEX-IS is free) servers. Can the
|> same be said for GLX/OpenGL?
There are two statements you make here. Let's examine each one closely.
The first is that because X terminal vendors are now introducing PEX terminals,
these terminals will make PEX computing less expensive than OpenGL computing.
Simply not true. There is nothing inherent in the GLX spec which makes it more
difficult or expensive to develop an OpenGL/GLX terminal. In fact,
knowledgeable technical types whose opinion I respect have made the argument
that less efort will be required to create an OpenGL/GLX terminal than a PEX
terminal. One terminal vendor has already licensed OpenGL, and I am actively
talking to a number of terminal vendors (and companies who make chips that go
into X terminals) who are seriously considering licensing OpenGL.
Some question whether the terminal model at present provides adequate
performance or differentiation versus a 3D workstation. The SHOGraphics PEX
terminals, whose 3D performance greatly exceeds NCD's or Tektronix's PEX products, are priced at a point not too distant from similar performing
workstations. If an end user has a choice between a 3D terminal or a 3D
workstation (which can serve all the functions of a terminal) for the same
price, most will choose a workstation.
The second assumption you made is that because the PEX-SI is 'free' on the MIT tape, this means PEX products will be less expensive. Untrue again. In fact,
as argued by Allen Akin in this space not too long ago, the PEX-SI is financed
by the dues vendors pay to the X Consortium (most large vendors pay $50K/year),
and by special assessments as required. For the first PEX-SI, Sun Microsystems
was paid about $750K by participating companies. Since a new SI will likely
be developed for 6.0, either the X Consortium will have to finance the effort,
or participating companies will either donate manpower or dollars to fund
it's development. There's no such thing as a 'free' SI (or lunch).
The OpenGL SI, which licensing companies pay an incremental $75K to receive, includes more functionality than the current PEX-SI. Functionality missing
from the SI needs to be developed by the implementor. When you consider the
volunteer effort contributed to develop the PEX spec and the SI, and the extra costs of implementing features not supported in the SI, and the special $ assessments, it is easy to see how those costs could exceed $75K. The per copy
royalty for OpenGL is a nominal $5. BTW, universities and non profit
organizations can get an OpenGL site license for $500.
The cost issue and the fact that PEX 5.1 products are released a few months
prior to OpenGL 1.0 products will not be the deciding issues on whether PEX or
OpenGL emerges as the more predominant standard. The more compelling issues I
feel will be ISV support, performance (both local and across the net), and long
term availability in key markets.
For a more detailed discussion of these issues, I suggest you read Allen Akin's
comparison paper of OpenGL/GLX and PEXlib/PEX. It was posted here about a month
ago. E mail me if you can't get to it, and I'll send you a copy.
