Quoting Chris Holmes <cholmes@anonymised.com>:
> > Yeah, it does, but I don't think it's for schema fragments. I feel that
_all_ schema information should go into AttributeType, the gt2 feature
model should grow to be rich enough to support it. And I guess your
original argument was that it'd be too hard, but I feel now that the
amount of work put into getting the dto's to work, and that will later be
put into enriching the gt2 feature model could have been saved by just
doing it now. But I understand that yyou guys aren't yet comfortable
enough to get that change through gt2, and I wasn't pushing hard enough,
to like actually drive it trhough myself. Hindsight is always 20/20.
Okay, we can proceed on that front. It would be nice to punt out gt2.0 first
since it would be the beginings of a API change. At the time I wanted to have
something that worked first, so I knew what to ask the gt2 side for. I think we
have that now.
> We would need to get the xml fragment into gt2 as well as min/max occurs
chris.
> Either that or keep a hashmap of fragment by attribupe name (ut I would
rather
> stay object oriented on this one).
I def. feel that anything that is a fragment should have a structure in
gt2. If we are to develop schema parsing in gt2 then we need a model that
rich. And honestly if you're doing a wfs client you probably should put
some good time into this, as gt2 is pretty weak in it's gml parsing. IanS
will help out, he just doesn't want to go it alone.
I have three months scheduled on creating a WFS DataStore, with my understanding
that such a beast would need to understand XMLSchema pretty damm well. My
current thought is to use the XML link idea for a first pass to reduce the
amount of XMLSchema to something manageable. Currently the GML3.0 gods are
doing this statically, the same solution could serve as dynamically.
Really hope that project goes ahead 
> Can I ask you to leave comments (like little bread crumbs), and I will ask
david to run through and answer any questions we run into.
Huh?
That particular class FeatureTypeInfo, has been where most of the *magic* goes
down. Not all of it is of David's making, Gabriel had a lot of magic in the
origional that took us a while to understand (I think it is the file that
triggered David's frustration with lack of comments).
And of course it appears to remain uncommented.
What I was trying to ask for was this: We can all leave TODO messages where we
are having trouble figuring out what is going on. And David/Gabriel can help
explain what is going on.