[RFC-24] Updated public draft

[Umberto Nicoletti]

Dear devs,
I have eventually updated rfc-24 to include the feedback collected
from the previous post.
In particular the following areas have changed:

1) the cache has been dropped because it requires a different
implementation for every supported mapscript. The reference counter
strategy has been adopted instead as proposed by Tamas. The various
mapscript objects will anyway need to be modified to keep a reference
to their parents to prevent early garbage collection and then
unexpected segfaults in the hosting language. This cannot be solved
with reference counting alone.

2) the various arrays of structures (layers, classe and styles) will
need to be changed to arrays of pointers to ease the development of
the new insert methods. This is also useful if int the future we plan
to remove the limitation of having a compile time fixed number of
layers, classes, styles, symbols, ecc

3) There is a tracker bug on bugzilla
(http://mapserver.gis.umn.edu/bugs/show_bug.cgi?id=2032) which has an
attachment that shows how I've implemented the previous item for the
layers array. This tracker will be used to post further comments and
code snippets

Sadly it seems that Ruby will still be left out of the game because of
how swig builds the extension code and will need a customized
implementation. Tcl needs some support from somebody that actually
knows it and moreover cares about supporting it.

The proposed plan is that, after this second discussion, I will
incorporate any further changes and then call for a vote on
mapserver-dev. If the rfc is adopted I will implement the changes at
the C level and in the swig interface files.
I will also take care of specific changes for Java and maybe Perl
mapscript. Someone else should volunteer to 'upgrade' the other
mapscripts.
I'll be doing this in my spare time and probably without funding, so I
cannot make promises about a time schedule.

I await your comments...

Umberto