(Present: Matthew, Gerard, Pat, Markus)

The purpose of this meeting should be to figure out what we're going to do/say at the interop.

Matthew: Gerard sent around a PDF with his take on utypes and their relation to VO-URP. We should now validate it against our use cases and requirements.

Gerard: VO-URP doesn't say too much about utypes usage per se, but the new document will.

Matthew: Omar and I will present something related to utypes to the TCG on sunday. The plenary session plans of last week still stand.

Gerard: I won't be able to present remotely, so Markus is going to have to talk.

Matthew: Last week we couldn't work out whether a DM is a graph or a tree.

Gerard: At least SimDM is a graph.

Matthew: Do we expect the other DMs to be non-tree graphs as well? If not, we should not overcomplicate matters just for the one case.

Markus: The DMs I've dealt with worked with trees, by virtue of them serializing into plain XML. Of course, they have implicit references, too.

Gerard: But making the references implicit in there explicit doesn't take that much engineering effort.

Matthew: When we have a tree, there's a straightforward way to make utypes parseable. And you have a straightforward way to create the utypes.

Gerard: But it's still much more useful to have a proper model in the first place. And with inheritance you have several types of relations, it's not only containment any more, so utypes syntax isn't expressive enough to cover all those cases.

Markus: Trees would also be nice because they're trivially traversable.

Gerard: In VO-URP, the model is just an XML document and thus a tree, so you can walk that. Most work appears to be figuring out what utypes on the various VOTable elements would be allowed to do.

Matthew: That's assuming (1) VOTable is the format we're considering and (2) VO-URP is acceptable to all.

Gerard: If you do a solution for VOTable, you can do it for other formats, too. Also, VO-URP has rather weak assumptions, so it's probably the leanest metamodelling framework we can use.

Matthew: Action Point on all tigers: Read and review Gerard's document.

Gerard: Timeline? The doc's not quite done yet.

Markus: I'd like to implant the question in your minds when reviewing: How do you like the paragraph on formatting utypes? I'd very much like it if "prefixes"/data model name (the thing in front of the colon) were fixed, so you don't need to keep track of prefix mappings to use utypes.

Gerard: Much of this is less about utypes but of serialization of DM fragments into VOTable elements. For that, it would be nice if we had references to tables and groups in VOTable.

Markus: This could probably work by just clients being aware that the value of certain utypes ("stc:AstroCoords.space.frame.reference") would be references to other elements; the mapping of ids to "elements" would happen in a serialization-specific way.

No telecon next week -- we'll meet again in Sao Paulo