Re: [Java-gnome-developer] Result of yesterdays IRC meeting

[Andrew C. <an...@op...>]

On Sun, 2006-04-30 at 21:06 -0400, Andrew Cagney wrote:
> - topic 3: autogeneration
> ??? from pygtk invited; very helpful on what they do;

That was Johan Dahlin - jdahlin. You can usually find him in
#gnome-hackers. Come to think of it, you can usually find most of us
there :)

>  My take is that 
> everyone agreed it was a good idea and that pygtk (even though it was 
> using !^*(@!)$^&* lisp) was going to be the best source for input files 
> as it had the most complete coverage;

The fact that it's in Lisp doesn't bother me per-se...

<anecdote>
Lisp never made sense to me, and the fact that many of the DocBook
formatting engines used Scheme markup to express stylesheets totally
bamboozled me. But about 18 months ago I had cause to learn XSLT and
although that is ALSO voodoo, it conveyed to me the idea of selecting
nodes and processing directives based on sets of nodes so isolated. Just
the other day I was reading the DocBook book 2.0 preprint and suddenly
the mushy DSSSL goo involved in selecting nodes and applying formatting
made sense to me. So I'm slightly less averse to Lisp than I once was.
</anecdote>

... but the fact that it is in Lisp is tertiary to the fact that they
seem to have formed a remarkably degree of completeness in their
articulation of the underlying GTK classes, functions, parameters and
return types.

I said tertiary because that in turn is secondary to the fact that it is
already done AND IN USE by another project. Assuming we were to go down
this road, to me the real win would be that we and the python bindings
could share the same set of meta data about GTK and friends. Yes this
would reduce overlap but I think even better it would improving the
number of eyes-on to the subsequent benefit of both projects.

It would have to be moved to a common module that could be in turn
depended on by both (for example) pyGTK and libgtk-java, but Johan
didn't think this was beyond the pale.

++

[There is still the huge issue of figuring out how to get our Java code
to align with such a theoretically generated JNI layer, but while not
completely orthogonal it is somewhat a separate question]

Follow up to java-gnome-hackers, I think.

AfC
Sydney