On Sat, 06 Feb 1999, you wrote:
Ted Harding, Chris Martin, Michael Johnson, Samy Elashmawy, zentara
have all brought forward relevant and interesting statements and points
of view in this thread.
However I think the single most important issue confronting the Linux
world is and remains: standardisation of a base Linux system.
There has to be agreement on things like file-system structure, the
init process, the basic libraries used and preferably also the naming and
the contents of the installation 'packages', the placement of these
libraries in the file-sytem tree, etc. etc..
If not it will become unnatractive for ISV's not specialized in Linux
to also make versions of their software available for Linux.
This is the same problem confronting companies who also develop for
Unix, you need seperate versions for DEC unix, HP unix, AIX, etc.
Even for companies specialized in Linux it will become irritating.
(this is the gist of the reaction of SAS when asked about a Linux
version of SAS)
And then think of the 'average Joe computer user' mentioned by Chris
Martin. Joe uses a SuSE 6.0 system and has discovered the ease of
working with rpm's. He discovers the rich source of rpm's available for
RH and downloads a couple and installs them on his SuSE sytem only
to discover that they either do not work or that he has messed up
his system.
(the fact that we also have other packaging systems for Debian and
Stampede is of less importance as long as tools like 'alien' are
kept up to date)
We should reach a point of agreement on basic structure of Linux
where it does not make one iota difference on which distro a rpm
was made.
[let me quickly add before Michael Johnson berates me: I too don't
think that rpms are the answer to everything. But they sure are
easy and comfortable for 'average Joe computer user' :-)].
I was already interested in this issue but this interest has
been reinforced by at least three 'incidents' this past year
where I was confronted with difference between SuSE en RH.
At the office I use RH 5.2 (glibc) and at home SuSE 5.3 (libc).
Everything I use at the office I also want at home, and vice versa
and the glibc-libc issue is the least of the problems.
My latest problem concerns the commercial statistical program Stata
that I use professionaly. Two weeks ago I received the new Stata 6.0
for Linux and promptly installed it at home and at the office.
[up to now Stata was no problem because it was aout, now they have gone ELF]
>From contacts I have had in the mean time with the developers I know
that the stata binary was compiled on a RH 5.0 system.
RH 5.0 still used libc5 so no problem you should think on SuSE 5.3, but:
1) after install there was all kinds of weird behaviour of the Stata
keyboard shortcuts and other things. However after some
messing about (I didn't change anything) and a couple of
reboots these problems automagically disappeared.
(and I thought hehe now I have a stable running system until see 2)
2) one problem that does not disappear is that in certain contexts
Stata without apparant reason crashes on SuSE (segmentation fault).
Stata is rock solid under RH.
(( To be fair Stata did give some problems under RH 5.2 but that was
because the developers had not checked under RH 5.1 or 5.2, namely
in RH 5.0 (and SuSE 5.3) 'xterm' is to be found under:
/usr/lib/terminfo/x/ while this was changed to /usr/share/terminfo/x
in RH 5.1 and later. A soft link solved this but there were probably
also some problems with ncurses which was also changed.))
I now have a binary compiled under RH 5.1 (glibc) and this runs
completely without problems under RH 5.2 and does not crash doing
the same things as under SuSE.
I spent days looking into these things and hours till very late last
night trying to make Stata also crash under RH.
So now I will have to document everything and send the data I was
using to Stata to see if they can reproduce the crashes under
RH 4.2 or 5.0 with the libc 5.0 Stata binary.
While I have to try the glibc Stata binary under SuSE 6.0 when I
get it.
These kinds of problems I can miss like toothache at the moment.
I you look at the number of Linux distributions available:
RH, SuSE, Debian, Pacific HiTech, Stampede, Mandrake, Slackware,
Delix, Caldera, Yggdrasil, etc. etc. this is going to become
a veritable headache for developers if these distros all start
embroidering on the 'base' Linux in order to differentiate.
Even ordinary testing under these distros would be a problem.
Or they will only develop for one 'standard' and that 'standard' is
going to become RedHat, like it already has at Stata.
-------------------------------------
Alexander Volovics
Dept of Methodology & Statistics
Maastricht University, Maastricht, NL
-------------------------------------
-
To get out of this list, please send email to [EMAIL PROTECTED] with
this text in its body: unsubscribe suse-linux-e
Check out the SuSE-FAQ at http://www.suse.com/Support/Doku/FAQ/ and the
archiv at http://www.suse.com/Mailinglists/suse-linux-e/index.html
