> On Mon, 31 May 1999, Julian Gilbey wrote:
>
> > Santiago Vila wrote:
> > > However, since every awk in the system is always a new awk and it is
> > > always available as awk, we could standarise the expectations and declare
> > > that every time a program in a Debian system needs any awk (either old or
> > > new), /usr/bin/awk should be used.
> > >
> > > What's wrong with trying to standarise things?
> > >
> > > In a Debian system, we have already standarised that perl should be in
> > > /usr/bin. We have already standarised that perl is perl version 5.
> > > Why can't we standarise the fact that /usr/bin/awk is a "new awk"?
> >
> > Yes, I see what you are saying, but why should we worry about tweaking
> > upstream software in various packages (and who knows which they'll end
> > up being?) to use "awk" instead of "nawk" when we can simply provide a
> > nawk -> awk symlink in every awk package?
>
> We could say the same for /bin/perl -> /usr/bin/perl, but we don't.
> Where is the difference?
The reason is when a script says something like:
#!/bin/sh
# Script to do blah
...
var=`cat $file | nawk '...' | wc -l`
...
then the absence of a /usr/bin/nawk will be quite serious. Do we know
whether there are any of those? I presume that Richard's list only
included the scripts which begin #!/usr/bin/nawk (correct me if I'm
wrong, Richard). It's one thing changing every #! line, it's quite
possibly a much more cumbersome job to check every piece of code to
see whether it calls nawk. The perl example is quite good as well: if
Debian were to decide that we were only going to have versioned perls,
so every script would have to begin #!/usr/bin/perl5 or
#!/usr/bin/perl5.005, then we would also have to modify every single
script which calls perl in any other way.
If you can check out every package on potato and convince me that the
only places where "nawk" is used is those five scripts and only a
handful of others, then I'll agree with you and say that we could
consider abandoning nawk. However, until then, I have no interest in
removing a harmless, and possibly necessary, symlink.
> The reason I asked about POSIX is the following:
>
> If we rely on /usr/bin/nawk being there, but POSIX said the only awk
> supposed to be in the system is /usr/bin/awk, then our packages would
> not be as POSIX compatible as they could be.
Huh? Then get rid of gawk as well, and mawk, and every non-strictly
POSIX-compliant variant of awk. You're being silly.
> > [...]
> > You're right, it would be nice to standardise every invocation of
> > (n)awk to actually call awk, but it is not particularly useful or
> > worthy of developers' time.
>
> Hmm, do you mean that better compliance to standards is not worthy of
> developers' time for five nawk scripts?
Compliance to what standards? "You may only have the following
software on your machine: MicroSoft Windows, MicroSoft Office, ...."
That's about the most common standard in existence at present ;)
Julian
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Julian Gilbey, Dept of Maths, QMW, Univ. of London. [EMAIL PROTECTED]
Debian GNU/Linux Developer, see http://www.debian.org/~jdg
