Minutes of the Utypes Telco / Wednesday 5 September CET 18:00

Participants: Omar Laurino, Gerard Lemson, Pat Dowler, Markus Demleitner, Matthew Graham, Mireille Louys

Matthew : reports on his survey about the current state and usage of Utypes in the VO. A first version of his note has been circulated in the Tiger Team. It points to some conflicts in the existing assertions found in various VO specifications. For instance :

• syntax of the utype string;
• parsable or not;
• Is a utype string unique among all the VO data models or not?
• String should be case sensitive or not ?

Mireille: Ambiguity on the terms explaining data model re-use, inheritance , data model extension , etc.. We should be more explicit about the context: we may inherit one class from one model to another one, and clearly identify for a utype string, which model or version it is refering to .<BR> Utypes, as labels do not inherit but the DM classes from one DM may inherit some classes from another model.

In order to resolve this, any time a DM re-uses a class , it should specify again all the data model fields in a new published document , and define its own list of utypes string, including the re-used ones .

This is to avoid circulating serializations concatenating too many data model definitions, where conflicts or ambiguities may not be easily resolved by the consuming applications

Name space : it is used to reflect which DM definition you are using .There is in fact a more general requirement, which is how can we uniquely bind a utype string to its related data model field and check , interpret its definition in the appropriate DM specification , and version. Whether it would be done by a name space mechanism / datamodel id/ string prefix/ url, etc. is an implementation decision once the requirement will be clearly expressed.

Meta Modeling / ( Omar and Gerard ? not sure )

XML schema, Utypes generation, and associated DM documentation can be generated within a metamodeling frame work, as proposed in VO-URP. Works well for complex data models like simulation DM. (see Gerard’s email)

Current usage: ( Matthew or Markus?) Get more from the DAL group. Check protocol implementations and the way they currently use Utypes. TAP schema defines a Utype column, that is mapped to a data model field in the ObsTAP spec for instance.

Markus suggests: try to sort out how we can prioritize the use cases listed on the tiger team wiki page. Vote for use-case to be supported using a table on the wiki for instance that every Tiger Team member could fill in according to his understanding and experience.

Markus: Utypes needed for STC (Coordinate System, etc.) when dealing with Coos in VOtable serializations.

Matthew suggests collaborative editing of his note on current usage.

Omar: which format for the documents: Latex ? HTML? Seems that HTML for intermediate reports is ok. Mireille prefers Latex for editing standards documents . Too much trouble with MS-Word.

Matthew will produce a tiger team report: on the Sunday TCG meeting in Sao Paulo and at the TCG telco tomorrow.

Gerard: not only flat serialisation to be considered: SIM DM needs several tables and the joints, as in the RDBMS relational architecture. Simulation data could be in any format, not necessarily FITS. (new invented format, HDF5, etc.)

Pat: Serialisating in table is always possible. But the hard part is to deserialise, and rebuild the hierarchical complex structure.

Omar: also several tables used for SED, not just one flat.

Pat: what do most people expect for Utypes? VOtable as the main serialisation formats. We should make clear that this is not the only covered use-case.

Next telco : Tuesday Sept 18th , CET 18:00 , other times , please fill in