Subject: Re: Linux beoming a real choice? From: hjstein@sunrise.huji.ac.il (Harvey J. Stein) Date: 7 Jun 1993 14:08:54
In article <C83rLr.3u6@jti.com> richb@jti.com (Rich Braun) writes:
A week ago, I downloaded Linux 0.99 patch level 10a. (I've been running
it and upgrading ever since 0.98 pl5.) This version contains SLIP, and
it also contains an interesting subdirectory:
ibcs
It only contains a code stub, but it got me pretty excited because it looks
like the developers are getting serious about the Intel Binary Compatibility
Standard.
This will allow Linux to run SysVr4 applications. And no, I don't think
this will compromise the design of Linux, other than to distract the
developers for a while during the process of implementing this compatibility
mode.
Speaking for myself, I think Linux won't become a serious contender (and
would remain a hacker's toy) until the range of applications is broadened
immensely. This is the fastest shortcut, barring Windows 3.1 compatibility
which I also hear is in the works (doesn't dosemu 0.49 run Win 3.1 apps
in "standard" mode?).
-rich
Rich speaks for me too! I'm currently setting up a 5 machine network
for my new company, and I would much prefer to run unix & X Windows on
the machines than to run DOS & MS Windows. And, of the available
unixes, I'd prefer to run Linux because I don't trust any of the
support here, so I'd have to support it myself (thus the source code
being available is important). Also, the prices are so ridiculous
(high prices for code that's mostly public domain), that I'll bend
over backwards to avoid dealing with them. Unfortunately, I'm going
to bend pretty far in order to get together all the applications that
we need. In particular, we need a database package, a development
environment for writing FORTRAN libraries (yes, I'm sorry to say,
FORTRAN) along with source code revision control, and statistical
packages. There are also a handful of DOS apps and MS Windows apps
that we'll have to run.
With linux, this means two things:
1. For the DOS & MS Windows apps, we'll need to have DOS partitions
with copies of MS Windows, and we'll have to reboot to get at
the applications.
2. For the unix apps, I'm going to have to stretch and strain to
find source code available versions which I can port to linux.
The first point is a pain, but can be worked around, and the second
means (probably) that we won't be able to use some of the better
products, and we might have to do without (does anyone know where I
can get the source code for an SQL server?). This also makes a
monstrous thing like Solaris much more appealing than it should be.
For example, for fortran development, I hear that there are some nice
integrated systems available for SysVr4, but if we use linux, we'll be
doing f2c | gcc. And if I can't get a symbolic debugger to work with
the original fortran source code, then we just won't be able to go
with linux. And even if this last point works out fine, we still
won't have as nice a development system as we could get if we were
running a SysVr4 binary compatible unix.
Even without dosemu being completely reliable, and even without some
WABI equivalent software for linux, what's going to stop me from being
able to use linux at work is the lack of SysVr4 binary compatibility, so
I hope that the ibcs directory cited above materializes fairly
quickly. As for the DOS parts, I think that dosemu will probably be
sufficiently robust sufficiently soon, and that sun will make the WABI
spec public, so linux will be WABI compliant, what, maybe a week after
that :)?
In spite of all this, I'm right at this very moment on disk 17 of the
full SLS distribution which I plan to install on my home machine
tonight. I'm going to check out what can be done with f2c and
symbolic debugging, and I'm going to have to start poking around for
database software.