rawhide report: 20050405 changes

[Dan Williams]

On Tue, 2005-04-05 at 11:27 -0400, Konstantin Ryabitsev wrote:

I have no personal stake in this debate.  That said...

On Tue, 2005-04-05 at 11:27 -0400, Konstantin Ryabitsev wrote: 
> It's more than pure "aesthetic considerations."
> I realize that this is very personal to you, since you are the author
> of libxml2, but I swear to you that my advice to switch from libxml2
> to cElementTree in Yum was based purely on merits and benefits for
> yum, and not on any personal likes or dislikes.

The point here is there is _already_ an XML parsing library in Fedora
Core and most other distributions: libxml2.  That library has existed
for quite some time as well.  We need to work on reducing duplicate
functionality, not increasing it.  This is not Debian with 12000+
packages, and they are unlikely to use yum at any point.  Fedora Core is
attempting to reduce duplicate functionality at this time, and adding
cElementTree doesn't help that goal.

If yum was already using libxml2, then it required more effort to switch
to cElementTree.  So why wasn't constructive work done to _improve_ the
Python bindings for libxml2?  Is there a list of items that need
improvement, and where are the bugs filed against libxml2 for each of
those items?

I don't think you considered the further implications of the choice to
move to cElementTree, simply the yum-centric implications.  Sometimes
FOSS is about working with existing projects to improve them, and not to
simply write your own solution when something doesn't quite work right.

That said, anything that makes yum faster is good, but perhaps working
with DV would help _all_ projects by improving libxml2, not just the
quick solution of switching to cElementTree just for yum.

Dan