Talk:Creole Open Issues

From LMNLWiki

I like the idea of the self-contained subsections.

I'm nervous about the attribute- and tag-ordering issues, but every time I try to think it through I find myself going down the tracks Jeni has laid, so I have nothing better to offer. I think these are basic areas of stress that Creole is just going to have ways of dealing with.

I'm afraid that if we expect much migration to occur between LMNL and nice XML (that is to say "classic" XML which has very limited capabilities to extend its model along the vector of annotations, given the designed-in weaknesss of XML), developers are just going to have to learn not to rely on annotation order, so as to preserve fidelity to an XML expression. Should we reconsider the decision to respect annotation ordering? IIRC that decision was made on the basis of "why not?" but we may now have found a reason. (And what's the use case for annotation ordering?)

OTOH I think structured annotations are too useful not to have, and that in itself impedes "reflection" in classic XML whenever they are used (which they will be). So maybe modeling with unordered unstructured annotations is just one of those things people who want to write their LMNL out in XML are just going to have to learn.

Basically LMNL as a superset is much wider, wilder and woolier world than XML, and not so much even from the overlapping thing but from the power of LMNL annotations. (I've thought for a long time that ordered structured annotations are the thing that will really set LMNL apart from other proposals in this space.) Once you're out there it may come as a surprise to some how tough it is to get back. I think that has less to say about LMNL than it does about XML, however (which I do love for its 80/20 simplicity and even its Procrustean character: when it fits at all, the bed is very comfortable).

--Wendell 11:06, 15 September 2006 (EDT)