Obscore minor updates: fixing STC data types re-use

[Markus Demleitner]

Dear DM list,

On Tue, Nov 19, 2013 at 07:48:55PM +0100, Louys Mireille wrote:
> this suggests minor changes in the names of a few data model fields
> inside Obscore.

I am completely neutral as to what these utypes look like; I *would*
have an issue with changing utypes *if* this meant a new version of
obscore (i.e., a new data model identifier), together with all the
problems that'd cause in terms of discovering ObsTAP services.

Therefore, I'd like to stipulate that if these changes end up in the
Obscore spec, this should be a "point release" (I believe that would
be version 1.01 according to SDP), and the data model identifier and
URI should remain "Obscore-1.0" and "ivo://ivoa.net/std/ObsCore-1.0",
respectively.

There is (IMHO sound) technical reasoning behind that, in that the
utypes aren't really used in the query part of the protocol (the
column names are), and no known clients depend on any utype being in
the response.  Thus, technically the utypes are part of the
documentation, and changing them will not break any part of a
client-server combo.

If people feel we need to bump either the model identifier or the
model URI, I'm against rolling out these changes.  The understanding of
old-style utypes has been that they're basically opaque, so we
shouldn't break clients and servers just to replace one opaque string
with another, even if we like the latter better.

Cheers,

        Markus