>>>> "ms" == Michael Sperber
<sperber(a)informatik.uni-tuebingen.de> writes:
ms> Could you summarize what problem you're trying to solve here?
ms> I'm a little lost.
There are three problems, actually. One is the path inversion between
source, especially "large" 3rd party packages, and installed
hierarchies. In source the structure is quite often
xemacs-packages/$PKG/lisp/*.el, while installed it must be
xemacs-packages/lisp/$PKG/*.el. This means that we can't run in place
in the package tree. Transitioning to $PKG/lisp in the installed tree
is a problem requiring some effort, I believe, because a number of
tools know about the structure of the installed tree, and we'd need to
check on which ones care about that.
A separate hierarchy offers a place to experiment with and make the
transition without screwing up old packages.
The second problem is "large complex packages". I especially have in
mind JDEE and Emacspeak, whose upstream maintainers won't touch the
XEmacs packaging because the installed hierarchy is too different from
their idiosyncratic ways of doing things. (42 JDE_DATA_GOES_HERE Make
variables, anyone?) Perhaps AUCTeX, too, but at least as of early
11.x it wasn't that big a deal to force AUCTeX into an XEmacs
installation.
Those packages already have their own idiosyncratic "find my info,
find my data" kinds of functions, so why not let them do it their own
way? The "opaque packages" proposes to enable just that by getting
the minimum amount of information needed, all of it hand-coded by
upstream (at least in principle), in auto-autoloads.
To complete the transition of making the source and installed
structures the same, we'd provide standard data-finder functions
(although mostly we have such functionality already, I suppose they
presume the existing structure, and also as far as I know do not
provide hooks for packages that have their own ideas about how to do
things) and of course XEmacs-maintained packages would have a nice
regular structure conforming to them.
The third problem is that the existing hierarchies have no good place
for executables and modules. Of course that's not quite right, the
real problem is that since they're architecture-dependent, we can't
distribute pre-built packages anyway. Still, lib-src and etc are
strange places to put scripts and the like; we'd like to have bin and
lib directories (and FHS fans probably want share :-).
--
School of Systems and Information Engineering
http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Ask not how you can "do" free software business;
ask what your business can "do for" free software.