>
> Apologies for the slow response.
>
> You have a good point and I agree that the coverage can, and in many
> case should be, approximate. The same for resolutions.
> However, I think there is merit in providing very precise information
> on the coordinate system that is in use, not to nail down the coverage
> are to a fraction of a mas, but to allow the client to assess whether
> it can actually handle the data (i.e., can it do the required
> coordinate transformations, or does the resource have the required
> precision). This becomes especially important when we start mixing
> with, for instance, the PDS.
PDS - planetary something?
You have an even better point - I am usually the one arguing that e.g. VLBI data also need milli-arcsec/ten m/s accuracy etc. and I had not thought through the implications of it being important for the the Registry to know how accurate/specialised a conversion service to call to execute a query (I think that is what you are saying?). I certainly agree that the information in the 'translated' schema fragment you posted will be used somewhere. I am still not sure if the most efficient way to do it it to make the registry that fine-grained. For example if people use keV as x-ray data spectral units, then every time I entery a query looking for x-ray data using metres as spectral units (which will happen more and more as the VO enables people to search unfamiliar domains) then every catalogue entry with spectral info has to be converted to m just to find out it is no use.
Also, in many cases the VO will simply provide an interface to the data centre's software to do specialised conversions.
It might be more efficient to use a single unit (or at most one or two) for each quantity, and make sure we store resolution in the S T & C domains in the basic Registry - and the information about specialised services needed, if any, in the Service Metadata section. The greater wealth of infomration available from the STC schema would be called upon only for the catalogues/data which were actually selected for further manipulation.
However I think that the only way we will find out what works best is by using it. The important thing is that if some parts of the IVOA start with registries which have simiplfied entries for coverage, the schemas can be extended if it proves necessary. Conversely if a schema based on STC in full detail is implimented and the data provider interface is manageable and the average 'local' VO can understand and maintain it, then I will happily eat my words and get on with it!
cheers
a