Hi Brian -
> How will a VO format for UV data be arrived at, ...
The approach taken in VO is very much like that taken in ALMA. A "VO format" for UV data really means one or more data models for UV data, plus a DAL service for each which implements the data model. The DAL service would allow UV data to be queried and accessed with active filtering and translation from whatever is actually stored in an archive. The service would define one or more alternative formats for retrieving the data, e.g., VOTable and any of a number of possible FITS formats. In addition to accessing UV data via a UV data model, one could have additional archive services which would provide access to this same data as an image cutout, 1D spectrum, and so forth, with the archive front-end performing the translation at access time.
For direct access to the UV data the first step would be to define one or more VO standard data models for UV data (e.g. for interferometric and single dish data, or conceivably one for both). All this would take is a small subgroup of people from the radio community to agree upon such a definition. It could then be moved through the IVOA data model working group to become a VO standard. It would need to conform to the IVOA data model standards (which we are still in the process of defining).
Anyone can participate in defining such a standard - it doesn't have to be someone active in the VO, although some VO folks would need to be involved. Of the people who are most active in the VO the first names that come to mind would be Anita Richards (Jodrell) and Jonathan McDowell (SAO/CXC, DM WG chair). Jonathan and I talked about possible radio data models only a few weeks ago so I know he is potentially interested. Radio and high energy data is similar in many respects and we would probably want to look at the similarity more carefully in defining a visibility data model for VO.
A key question is what is a useful data model for UV data. In general VO data models are defined only for fully calibrated data. One would like a VO data model for radio data to at least be telescope independent: telescope-specific information could be included in actual datasets but the core data model and format would be telescope independent. There is no reason we couldn't define a VO data model for raw UV data so long as it was applicable to data from most radio telescopes, and the VO would be the logical place to define such a standard. However, it might be more useful to define a UV data model for calibrated data, and apply the calibration on-the-fly on the archive server when accessing data from an archive via a VO service. How feasible this is, and which of the alternatives would be most useful is not clear.
The ALMA data model is currently too telescope specific to be a VO standard. However in general form it is already largely telescope independent. If it could be structured as a generic core data model with a mechanism for associating the telescope-specific information required for calibration, then it could provide a VO standard for raw interferometric UV data. Or we could define a standard for calibrated UV data (if such is found to be useful - I am not sure of this) and possibly include in the project a repackaging of the pipeline to provide a service to serve calibrated UV data.
> ... and should ALMA play a role in this process (i.e., can it be tied to
> our "data model")?
I think ALMA should consider helping to promote this, particularly considering the timing, with ALMA and VO defining similar data models more or less at the same time. Integrating the ALMA archive with the VO is already a goal of the project and this is one of the most important problems to be solved to do this.