Distro non-wars... was Re: [plug] Red Hat

Denis Brown dsbrown at cyllene.uwa.edu.au
Tue Sep 23 09:29:09 WST 2003


Thanks, James.

On Mon, 22 Sep 2003, James Devenish wrote:

<snip>

> > But when all is said and done, could I not take a native "RedHat"
> > installation and by dint of swapping one desktop for another, one set of
> > wallpaper for another, tweaking the security, turning off (or adding)
> > services, etc, etc turn it into a Debian installation as far as look and
> > feel is concerned.
>
> It will be harder than you might be currently imagining.
>
Having read below, I see what you mean!

> > Others in past threads have also eluded to specific distributions being
> > preferred because of similar "constraints" from other software.   Pardon
> > my naivety, but is this really a limitation?   Aren't we really talking
> > about "hooks" into the OS such as library revision levels, which need not
> > be the province of a specific OS vendor or flavour?   Ergo it should not
> > matter which distro a developer "prefers" - they should all be capable of
> > working??????
>
> You might be surprised. It *is* possible to produce products that will
> work under arbitrary Linux distributions. However, it requires actual
> care on the part of developers. In particular, there are areas where
> these is simply no consensus. When we talk of "distrbutions" we usually
> mean "operating systems" -- although the kernel and GNU utils are common
> to all distributions, there is a vast realm of issues for which there is
> no canonical configuration. For example, different distributions apply
> their own patches to software; different distributions write their own
> manual pages. Some, like RedHat, have a particular penchant for bad
> default settings (hmmm,...I guess I've been burnt more than a few times
> by RedHat!). Just look at something like Debian policy documents to
> realise the vast decision-making that every distribution makes. Don't
> mention glibc. This is not to say there's anything "wrong" with this
> situation per se -- the situation exists because everyone has different
> needs. On the other hand, there are some "reliable" things such as
> POSIX, X11, gcc, and so forth which make it much more practical than
> porting to, say, Windows (and clearly porting to Windows is possible for
> a lot of software). However, just look at this:
> <http://www.linux-debian.de/openoffice/>. There's probably no reason why
> any software for one free Linux OS (e.g. RedHat) could not be supported
> on another free Linux OS (e.g. Debian) but it could very well take
> effort. On the other hand, it might just work the first time  The
> problem for commercial products is that although your software might run
> absolutely perfectly on all OSes, you can't know this for certin unless
> you actually try it.
>

As you say, the situation looks a lot more challenging than I had thought.
Basically, if one created (or in user-land, downladed) a binary it would
be pretty much locked into a specific distro, or at the least, into a
specific set of underlying libraries and APIs.

I now understand more of what Jon was saying (and have had the benefit of
Jon's reply to this thread also; thanks, Jon.)  I'll have a look at the
references James mentioned and spend some time looking into Linux
standards generally.

Thanks for your time,
Denis


_______________________________________________
plug mailing list
plug at plug.linux.org.au
http://mail.plug.linux.org.au/cgi-bin/mailman/listinfo/plug


More information about the plug mailing list