Hi all,

A useful question to consider is this: what _could_ be done, but shouldn't?

(1) GUI frameworks.  These are usually somewhat painful to build (e.g.
Lua-Gnome) and then it's best to leave it to the package manager (e.g.
Debian liblua5.1-gnome-0 does the trick). However, not everyone has
the luxury of a package manager, or one that contains the desired
package.  And LuaRocks cannot be expected to make fine distinctions
between various flavours of Linux, or to intuit that an OS X user is
using MacPorts or whatever.

However, it would be nice if things like IUP Lua or wxLua was
available as rocks...

(2) Linux and OS X binary rocks. Many think that this is a complete
no-starter, probably because libc hell made binary compatibility
awkward. The situation is now much better and maybe some kind soul
could make binary rocks available on some repo somewhere. The argument
is not that Linux people are afraid of building, it is that some rocks
have complicated build dependencies.  However, what happens when a
binary fails? The situation would be clearer if some minimal
auto-testing were done.  Even Windows is a binary tarpit, because of
the runtime dependencies. Sometimes Ryan (from Lua for Windows) gets
mad at the weird MS toolchain and threatens to move LfW over to mingw
- that would certainly help!

(3) LuaRocks as a distribution delivery system. Last year I
experimented with Lua for Linux, which had a captured LuaRocks
installation. Now I think the best way would be to construct a
meta-package which pulls in all the desired packages that make up the
distribution - a Rock Pile.  Using post-install hooks, this could do
all kinds of cool things. For instance, SciTE can be taught to debug
Lua (on GTK+ platforms and Windows) and the necessary hooks could be
made after installation - ditto for gdb (I hope to make both
scite-debug and luagdb available as rocks soon)   At least on Gnome,
one could register .lua as a file type and associate appropriate
actions.

(4) Would it be useful/appropriate to have independent implementations
of Lua such as Metalua and LuaJIT available as rocks?

It's interesting to see how far a OS-independent packaging system can
be pushed, and it's useful to think about where it should be pushed.

steve d.

_______________________________________________
Luarocks-developers mailing list
[email protected]
http://lists.luaforge.net/cgi-bin/mailman/listinfo/luarocks-developers

Reply via email to