Linux Today: Linux News On Internet Time.

Is the Server market that based their OS's on AT&T Unix really fragmented?

Aug 28, 1999, 15:54 (9 Talkback[s])
(Other stories by Brent Metzler)

Contributed by Linux Today reader Brent Metzler

A bunch of companies developed hardware and needed an OS. They had 2 choices. They could build an OS from scratch that would be "unique" or they could build an OS from an already strong sourcebase. In either case they ended up with a competely new OS, it didn't really matter where they started from. So the fact that they all started from the same thing doesn't make the products fragmented.

This is where people "miss" it. Solaris and IRIX are in practical terms as different as Solaris and NT. They aren't intended to be fully compatible. When I look at the Mozilla source, I see code specific to Windows, Linux, Mac, BeOS, and OS/2. I could very well say that the desktop computer market is fragmented, with more evidence than the Unix/Server market. The Unix/Server market is cognate, not intended to all be identical.

You say that you can compile NT and it works on Intel and Alpha. This is not a good analogy to compare to, say, Oracle running on multiple server platforms. NT is NT no matter what chip it runs on. NT will never be OS/2, or Mac, or anything else. Solaris now runs on Sparc and Intel chips. Compare the ease of recompiling Solaris apps for sparc and Intel, and NT for Intel and Alpha. Although I have no experience I am confirming with a friend the differences with Solaris on Sparc and Intel. I would seriously be surprised if there are any.

Similar does not mean bad. It does not mean a carbon copy. I see it as a good thing that the companies choose to base their OS's off of a common beginning. Each company has a strength that the other companies don't. By being similar, and not different like the desktop market it provides a greater ease to porting software to multiple platforms.

Wanting even greater uniformity in the server market, the Posix standard was born. The Posix standard provides a common framework across multiple OS's. This is not an acknowledgement that the server market was somehow "fragmented" before, but a continuing innovation in the value that the server market provides. As far as being fragmented over the Posix standard, which is a valid concern, which OS's do not support POSIX.1 right now? Okay, so NT does pay lip service to POSIX.1, but for just one example of what this means, check out what CAI had to say about porting their application to NT. Because Microsoft wants to compete in the server market, I would say that this lack is a fragmentation of the server market. I guess that the question should be, if there is a fragment in the market, which side should you choose? The side that fragments to keep market share, or the side that decides to provide value and real innovation to the customer.

What about fragmentation of a single OS? When I write an application for an OS, I think that I'd like it to be able to write it for the OS and have it run on all the various incarnations of that OS. For instance, I am going to use Linux here because it has been ported to the largest number of different hardward platforms. Linux will run from the smallest embedded systems, all the way up to 64bit chips such as the SPARC and the Alpha. I can be reasonably confident that *unoptimised* code I write for Linux on one hardware platform will probably also compile/run on another hardware platform. Windows also runs on different hardware, with different implementations of the OS. An Application I write for NT, may not work on 98 or CE. It may work, but probably require at least a little porting. I'd consider this fragmentation of an OS. When Windows 2000 is finally release Microsoft is rumored to have 9 implementations of the OS which will replace the diverse OS's that Microsoft currently supports. Will we see an end to fragmentation inside the Windows OS at that time? I seriously doubt it. I consider the failure to write a single piece of code that compiles over the implementations of one OS more severe then code that doesn't compile on more then on OS, for instance, Windows and Mac, or NT and AIX, or Unixware and Solaris, although they last 2 have the greatest possibility of compiling identical code. Does the fact that they have a greater chance of compiling identical code have any meaning as to whether they are more fragmented then the other combinations?

In conclusion, what am I trying to say here: Microsoft spends a lot of money on marketing their products. They also are seen as a credible source. So they can say things that aren't true and people will accept it without researching the facts. Moreover, there are many people who are not able to perhaps understand what is being talked about and has to take whatever is said as being fact. Perhaps these are doctors, or lawyers, or just plain old managers. They may be the smartest people in the world, but it's not their job to know about the server market. Microsoft says that they bring value to the server market by having a unified server OS, while the rest of the market is greatly fragmented. I hope to have shown in this document that this is false and that the server market is basically cognate, and not therefore not fragmented, but actually has a standard for providing a unified interface. It was also my desire to show that it is really Windows that is fragmented in the server market, and not only fragmented in the server market, but also fragmented in it own implementations.

-- This is the first part of (hopefully) an ongoing work. (C) Copyright 1999 Brent Metzler