>>>> "dv" == Didier Verna
<verna(a)inf.enst.fr> writes:
dv> sperber(a)informatik.uni-tuebingen.de (Michael Sperber [Mr. Preprocessor]) writes:
> Didier, I think there's a big misunderstanding here: Sure, I
disagree
> with you. But that's mainly because our software installation setup
> is different from yours, and the present scheme fits ours pretty well.
dv> [...] You'd understood that it doesn't fit mine.
Ah! No, I hadn't. Please explain why.
dv> That's disapointing as well as worrying to hear that. It sounds like
dv> you're closed to any discussion on this just because it is a problem (or a
dv> feature) you just don't want to care about.
I don't even see your problem yet.
I think it's wrong to intermingle the design issues concerning the
package layout with the global issues of XEmacs directory layout. If
you make one depend on the other, it's much harder to make local
changes. I want to be working on the package system, not the global
directory layout. For the directory layout issues, I don't stand a
chance against you, and I don't want to.
Moreover (and please acknowledge this, Didier), there are software
setups *other* than yours (notably, ours) where hard-wiring what
you're suggesting is going to royally fuck things up.
At the file system level, a package constitutes a collection of files
that belong together. With hierarchical file systems, the *only* way
to reflect "togetherness" (as well as "apartness" from other
packages,
say) is to put those files into a common subdirectory which is not a
subdirectory containing other packages. This has a number of
advantages, two being ease of installation and un-installation as well
as an easy-to-satisfy integrity criterion. The latter point is
especially important since I can check package integrity *locally*,
and it's pretty hard to violate it without seeing what you're doing.
Once the package is spread out as you suggest, installation is still
easy enough, but suddenly, if I want to uninstall or move or ship a
package, I have to remember where all its bits are. If I forget some
of them, there's no guarantee the system will *ever* notice that fact.
dv> Again, I'm sorry, but no way. This is like doing consciously something
dv> wrong first, and then reparing the damage afterwards. I'm not ready to make
dv> dozens of symlinks even if it's automatized in any fashion.
Well, so there. I'm not ready to spread packages over disparate
portions of the hierarchy.
Seriously, links *are* Unix's native mechanism for expressing sharing.
You may not like it, but it's the only thing there is. They're not
exactly a neat design, but they're excruciatingly easy to manage
automatically.
Yet another solution for achieving maximum sharing without links is to
factor the architecture-specific portions of a larger packages into
packages of their own. This makes sense for staging reasons as well.
Stick the architecture-independent bit into a shared package
hierarchy, and the architecture-specific bit into one conditionalized
on architecture.
dv> But please, don't say things like "I won't change it".
I never said that. How come you claim this?
--
Cheers =8-} Chipsy
Friede, Völkerverständigung und überhaupt blabla