Subject: Re: Why is TCL_SHARED_BUILD not 1 in the recent tcl80 package?
To: None <firstname.lastname@example.org>
From: Greg A. Woods <email@example.com>
Date: 01/09/1999 13:33:26
[ On Sat, January 9, 1999 at 04:36:17 (+0100), Hubert Feyrer wrote: ]
> Subject: Re: Why is TCL_SHARED_BUILD not 1 in the recent tcl80 package?
> On Fri, 8 Jan 1999, Thorsten Frueauf wrote:
> > fix the postgresql PLIST accordingly. While there it should be
> > decided whether its a good idea to link tclsh8.0 to tclsh and
> > wish8.0 to wish. I don't know whats the right thing here either.
> > (This is pointed out in pr 6757).
> Well, at least the perl pkg does so, and I think it wouldn't be bad to
> call things by their name without version:
I sorta like the idea that was forwarded to me in private mail, i.e. to
provide a meta-package that installs and owns the symlinks that point
the generic name to the one with the version number. One can then use
the pkgtools to swap the default version of a package. This makes some
things easier too, such as managing key files between versions of SSH
where a procedure must be run to "do the right thing" (though SSH is a
bad example because it, like tcl and perl, has some degree of
understanding within itself of version interdepencies).
However there's something slightly in-elegant and incorrect about this
whole mess. Unfortunately the only proper SCM solutions I can think of
involve a whole lot more complexity and maintenance. Ideally package
authors should think of these issue and provide the solutions, perhaps
ones that are unique to their software. The problem is that software
authors often have an entirely different mindset about these problems.
Greg A. Woods
+1 416 218-0098 VE3TCP <firstname.lastname@example.org> <robohack!woods>
Planix, Inc. <email@example.com>; Secrets of the Weird <firstname.lastname@example.org>