Re: Pango branched

[Owen Taylor <otaylor redhat com>]

Daniel Veillard <veillard redhat com> writes:

> On Sat, Apr 13, 2002 at 11:28:09AM -0400, Owen Taylor wrote:
> > 
> > I've made a pango-1-0 branch of Pango to separate out the
> > GNOME-2.0 version from pango-1.2 changes (such as a switch
> > to Xft2) that I expect to start checking in soon.
> 
>   Hum, yes it's time to start announcing plans for the evolution
> of the core libraries.
> 
>  I have been working on XML Schemas support for libxml2 for the
> last month but didn't commited it to CVS because it was really not
> in a shape to be pushed. Things are starting to take shape and I
> would really like to push to CVS. But I'm not fond at all of branching
> last time I did was 2 years ago and I still have the sequels from
> years waiting for the Gnome project to resync. 

I would like to point out that what I'm doing for Pango
is fundamentally different from the libxml1 / libxml2 split.

There is no danger that GNOME will be stuck on an old version
of Pango since no incompatible changes are being made.
All the changes that are going to be made to Pango are going
to be binary and API compatible... the branch is there:

 - To make sure that we don't destabilize what people are
   testing against.
   
 - So we can make maintainence 1.0.x relesae

 - So that people won't accidentally use features with
   a not-yet-stable API.

I won't comment on whether it makes sense to branch libXML or
not ... the question there is really just whether you think
there is a chance you'll need to do a maintainence release
without the new API before you feel comfortable shipping
the new code. (And whether you think you can do the changes
without destabilizing the old parts of the API.)

>   Since the change are basically adding new modules in the XML Library
> (unicode, regexp and schemas) and not modifying any existing interfaces
> I'm really tempted to keep the head being the current libxml2 version
> and at least in a first time not have Schemas support compiled in by
> default (i.e. configure will not select it unless asked). However

What we did for Pango is that unstable parts of the API rae compiled
in by default (since they were are used internally among other things)
but you have to make a special #define to use them.

Regards,
                                        Owen