Subject: Re: tcl package names....
To: None <firstname.lastname@example.org>
From: Alistair G. Crooks <email@example.com>
Date: 01/07/2000 02:25:15
> * Hubert Feyrer (firstname.lastname@example.org) [000107 09:39]:
> | On Thu, 6 Jan 2000, Matthias Drochner wrote:
> | > Added Files:
> | > pkgsrc/lang/tclman80/pkg: COMMENT DESCR PLIST
> | IMHO tcl80 and tcl80-man would have been more sane.
> | (or even better: tcl80-bin, tcl80-man, and a tcl80 that pulls them both
> | in)
> I guess the idea behind that was, that you can use a simple pattern
> to match tclman80, tclman82 and in February tclman83. The same applies
> to tk, where we might run into problems, if the match would be something
> like tk*-man and tk*-bin. Just image a package tkxyz, that gets split
> into tkxyz-man and tkxyz-bin. Suddenly the match, that was supposed for
> major versions, applies to the xyz name as well.
Why is it necessary to put the version number into the directory
names? For most of the time in pkgsrc, we have managed to get by,
quite nicely, with pkgsrc/<category>/<package> style names.
Occasionally, when there's new functionality, but possibly unstable,
in a later, or beta release, then we use
pkgsrc/<category>/<package>-current in addition.
Now I'm well aware that tcl and tk do not fit into the example
given above - tcl 8.2 has been out for a long while (August 1999),
and is stable.
So why hasn't the tcl80 directory been nuked, along with the
tcl-8.0.5 package, and a new tcl directory been created to house
the new tcl-8.2.3 package?
Why are two packages necessary? It's not as if there are the pains
involved in the jump to the bytecode compiler in 8.0, or the (fairly
radical tk 4.2 -> tk 8.0 jump).
(The tcl80 package name, and then tk80, came about because of idiocy
on my part, I'll admit that. I'd just like to clean it up now, and
it looks like we're just getting into a bigger mess).
Alistair, aka Confused of the UK.