On Wed, 2005-07-20 at 22:41 -0700, Bryce Harrington wrote:
On Wed, Jul 20, 2005 at 02:49:18PM -0600, John Taber wrote:
I think this is a much bigger issue than the current Inkscape release - but maybe some here are in position to help coordinate between the differnet gtk family groups. There are more than a few of us dealing with this - it seems like over 10% of the posts on the gtkmm developers list are about library/installation issues and it also seems to be a big issue for Inkscape release packaging. The problem stems from many independent library dependencies, different releases, and different version nos.
Windows seems to have managed the shared dll hell problem by having very, very few, and more carefully controlled/tested releases, however, that seems to be a less desirable approach than frequent upgrades. Somehow I think a more unified gtk/gtkmm "studio" release approach is needed so that all the major dependency libs are "packaged" (even more than what Debian does) including glib, gtk, pango, libsigc, libglade, gtkmm, etc, etc.
Yes, I definitely agree. After all, we're application developers first, package distributors a distant second. We need to be looking for ways to move that responsibility upstream, as we've been able to do with the unix packages. We used to put great attention into crafting debian, gentoo, rpm, etc. packages each release for each distro, but as time has passed and as we've gotten into more and more distros, today we are able to mostly leave that work to the experts and focusing on just putting out a good core package. Being able to take that same approach with the Windows package would sure simplify things; consider how much trouble the Windows build has been this time through, and the extra work required for testing and fixing it up. If another group were able to take the responsibility for this, it would simplify our release process and enable the Inkscape releasers to focus even more on the software, rather than the packaging.
Another thing that I have noticed here at desktopcon.org is how many times developers talked about shifting generalizable solutions upstream. Carl Worth mentioned this much about his style of dev. with Cairo. This really makes sense to me in this case as well. (CREATE project anyone?)
Jon