From: jliddle@rs6000.cmp.ilstu.edu (Jean Liddle) Subject: Re: Linux vs. the world Date: Tue, 11 Aug 1992 03:30:29 GMT
In article <1992Aug10.180715.789@crd.ge.com> davidsen@crd.ge.com (bill davidsen) writes:
>In article <1992Aug08.001344.15942@rs6000.cmp.ilstu.edu>, jliddle@rs6000.cmp.ilstu.edu (Jean Liddle) writes:
>
>| In what way? When I was working at Bayer last summer one group was using
>| SCO unix. The compiler couldn't even handle floating point numbers
>| correctly, and unless you had purchased support, no free fixes were
>| forthcomming. Furthermore, the system crashed more than once/day.
>
> Obviously your system had been seriously trashed by someone at
>install, because f.p. has worked on every version of SCO back to the 286
The floating point bug had been documented -- I became involved becouse
the department was fed up with dealing with the German branch of the
company, and wanted someone (me) who spoke fluent english to handle the
contact with SCO. We were told to deal with the European subsidiary,
who would not release any fixes without a support agreement. The
problem had to do with inconsistencies in how SCO dealt with , vs .
as a decimal delimiter -- one part of the code only accepted .'s, while
another only ,'s. I left at the end of the summer to return to school,
so I do not know the final outcome. The people working on the SCO machines
were capable unix people (familiar with sys V and VAXen and, to a lesser
degree, with BSD and Suns).
> Based on over 100 machine-years experience with SCO I don't believe
>there is a reliability problem in the product. Based on maintaining
>almost 300k lines of production code on multiple machines, I don't
>believe there's a problem with floating point. Based on having gotten
>many fixes from sosco and uunet by uucp and ftp, without paying any
>support cost, I don't believe there is any lack of support for all SCO
>customers.
>
The American portion of the company may have different marketing/support
policies that the European/German branch. I don't know. I can only
relate the once experience I was involved in, after which I would never
touch an SCO product. This, of course, is only my opinion.
>well have a hardware problem or have been seriously misconfigured. That
>doesn't make SCO bad.
>
There may have been configuration problems which led to the frequent
crashes (I have no way of knowing). However, the original installation
was performed by an SCO representative (or reseller of their product).
The floating point problem was a flaw in the compiler and how it
interacted with the rest of the os. As this was not my project, I am
not in a position to say a great deal more about it.
Jean.
-- Jean Liddle | == Bill Clinton for President == Computer Science, Illinois State University | e-mail: jliddle@ilstu.edu | - Carol Mosley Braun for Senat -