Xref: utzoo comp.sys.ibm.pc:38548 comp.dcom.lans:3840
Path: utzoo!attcan!ncrcan!becker!geac!jtsv16!uunet!samsung!ctrsol!cica!iuvax!purdue!bu-cs!xylogics!world!madd
From: madd@world.std.com (jim frost)
Newsgroups: comp.sys.ibm.pc,comp.dcom.lans
Subject: Re: Xwindows & PM (OS/2)
Message-ID: <1989Nov17.172414.21727@world.std.com>
Date: 17 Nov 89 17:24:14 GMT
References: <2528@umbc3.UMBC.EDU>
Distribution: usa
Organization: Software Tool & Die
Lines: 57

battle@umbc3.UMBC.EDU (Rick) writes:
>My boss hit me with a question today that was a surprise.  He
>wanted to know what the relationship is (if any) between Xwindows
>and PM.  I have never seen any software vendor who writes for the
>DOS-OS/2 (PM) world also write for Xwindows.
> 
>My feeling is that Xwindows will stay in the UNIX world and PM
>will stay in the OS/2 world.

I think it's more likely that X will be a large multivendor product,
of which PCs are a subset, while PM will remain exclusively on PCs.

There are some differences between PM and X, but generally X is more
advanced.  It has support for a wide range of hardware types, while PM
is pretty narrow (and simpler because of this).  Both have design
flaws which make them a pain in the neck for developers, but they're
different flaws.

Several software houses with which I am familiar have chosen the same
path I have in the past: implement a library layer which your
application speaks to, and re-implement that library for each new
environment.  This layered approach works incredibly well.  I'm fairly
certain, for instance, that I could move several applications I've
written to PM in a few days including the learning curve.  In one case
we moved an application from X11 to SGI GL in four hours (they are
extremely dissimilar) with this kind of architecture.  Not bad for an
intensive graphics program with better than 35k lines of code.

The biggest reason why you haven't seen a lot of companies do this for
their products is purely audience.  Traditionally PC-based companies
are sticking with PCs and porting to PM, since their traditional
customers have PCs.  Traditionally workstation-based companies are
sticking with workstations, which means X right now, for the same
reason.  I think we'll see more merging as PC vendors realize that
they can get more done easier on workstations than on PCs, but I don't
expect workstation-based vendors to go to PCs -- it's too hard to get
complex applications to run (you don't want a 40,000-line application
to be written in MSC, let me tell you, even if people had PCs which
could run such a beast, and 40,000-line applications are low-end on
workstations) and the people that want them just don't want to deal
with PC problems.

There are other reasons why vendors aren't writing for both.  The
programming environments are quite dissimilar.  I heard rumors that
the OS/2 people deliberately avoided UNIXisms when designing their
product.  This strikes me as asinine (there are many things in UNIX
which are done cleaner than they are in OS/2) and it makes it
difficult to get good portability between OS/2 and traditional X
environments.  It works the other way, too -- threaded designs don't
port to most UNIXs right now.  But then again, most programmers don't
use threaded designs effectively so it's pretty moot.

Food for thought.

jim frost
software tool & die
madd@std.com