Ville Skyttä writes:
Given the low number of available modules currently, I think the flat
"make
install" (ie. all modules installed to one dir, no subdirs) is fine.
I think this needs thinking about. First, that means that
run-in-place and installed modules have to keep working differently,
and that's the source of the current problem.
Second, I hope to do something about the module population. I have
packaged up libneon and libcurl, which are just waiting for an
application and some refinement of a common API to go public (that's
three modules there). I'd like to do the same as canna with the wnn
and sj3 input methods. zlib and base64 should be actually done as
modules. It's possible that all the Mule codecs could be moved into
modules. (Lawd, lawd, I'd *love* to get that crap out of core!)
Third, some of the modules really ought to come with Lisp code (eg,
higher level UI for neon and curl).
BTW, without knowing much at all about the module search code and
even less
about it on eg. Windows; does loading a hardcoded thing like "foo/bar" with
the forward slash path delimiter work on systems whose path delimiter is
something else?
Ah, good point! I'll try to fix that. (Yes, it should work on
Windows; in fact Windows will accept either convention AFAIK. The
historical problem with / is that it is the MSDOS option character.)
_______________________________________________
XEmacs-Patches mailing list
XEmacs-Patches(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-patches