TAPRegExt erratum, Identifiers for Obscore

[Mark Taylor]

On Wed, 4 Dec 2013, Markus Demleitner wrote:

> Dear DAL list, dear DM list,
> 
> There are various ways clients can discover obscore services (or, for
> that matter, TAP services containing other data models); the ones I'd
> propagate are looking for the data model's ivoid and/or name.  This
> information is communicated to the Registries in the TAPRegExt
> capability.
> 
> And here the trouble starts.  First, I messed up.  For some reason I
> cannot find, a bad identifier sneaked into the example in the
> TAPRegExt draft; it lists ivo://ivoa.net/std/obscore-1.0 as the

>From the department of the super-picky: as far as I can see, 
TAPRegExt says "ivo://ivoa.net/std/ObsCore-1.0"
(different capitalisation).  Since they are both wrong it obviously
doesn't make a big heap of difference, but for anybody doing forensic
debugging to work out where certain mistakes crept in (possibly me
at some point) it's useful to know exactly what's wrong and how.

> standard id for ObsCore.  Unfortunately, ever since ObsCore did state
> their id, it's been ivo://ivoa.net/std/obscore/v1.0.

Evidently I copied TAPRegExt and not ObsCore when implementing taplint
(the TAP verifier), so the released version will only identify ObsCore
services as ones that report a DM of "ivo://ivoa.net/std/ObsCore-1.0".

I've fixed it for future versions so that it recognises both variants,
but reports a warning if the wrong one is used.  This will be in the
next STILTS release, pre-release available at
ftp://andromeda.star.bris.ac.uk/pub/star/stilts/pre/

> Now, of 7 services currently registering ObsCore services in the VO,
> 6 use the wrong one (Kudos to the operators of ivo://jvo/subaru/spcam
> who got it right).  That's bad, and I deserve some spanking.
> 
> So:
> 
> (1) If you have an obscore service, please make sure your standardId
> is ivo://ivoa.net/std/obscore/v1.0 rather than whatever is written in
> the current TAPRegExt example.
> 
> (2) How can I clean up the TAPRegExt example as quickly as possible,
> before even more implementors get confused?  Given that this is not
> normative text on TAPRegExt's side, could we do some sort of fast
> track?  Or can we add an erratum to the document entry page while
> matters take their proper way?

I think this is a wider issue.  There is currently no lightweight
way of flagging up errata, or more broadly things which shouldn't
have been written into standards and need some further explanation,
for IVOA documents.  The only thing you can do is go through the
whole process of issuing a new version.

It would be nice to have an erratum page associated with each IVOA
standard.  This could probably just be a wiki page, but should contain
things which are known to be problematic and possibly workarounds.
It would serve two purposes: first, act as a list of things that have
to get fixed in the next version, and second, as a reference for
implementors etc who are looking at the standard and scratching
their heads wondering how the text can possibly make sense;
often such things are known within the relevant VO (sub-)community
but there's nowhere to record them.

> Second, there's a related issue; I (probably foolishly) choose to
> include the concept of the data model name (in addition to its ivoid)
> in the TAPRegExt document as well.  Now, it's always a bad idea to
> keep the same information in two places and in two forms, and so
> that's the case here, too; check this out:
> 
> http://dc.zah.uni-heidelberg.de/__system__/adql/query/form?__nevow_form__=genForm&query=SELECT%20ivoid%2C%20detail_value%20FROM%20%0A%09rr.res_detail%0AWHERE%20detail_xpath%3D%27%2Fcapability%2FdataModel%27%0A%09AND%201%3Divo_nocasematch(detail_value%2C%20%27%25obscore%25%27)%0A&_TIMEOUT=5&_FORMAT=HTML&submit=Go
> 
> So -- some (most) services have "ObsCore-1.0", some "Obscore 1.0".
> The forms with the dash surprised me a bit because
> 
> (a) the TAPRegExt example has the form without the dash, and
> (b) the form with the dash is not present in the ObsCore spec itself;
> it gives the dash-less form in the only place that could be construed
> as a self-christening.
> 
> 
> And yet everyone but myself uses the dashed form.
> 
> So: what's the correct form?  Is there one?  What about
> capitalization?  Is there some convention on DM names that I'm not
> aware of?  And how do we make sure that in the future everyone uses
> the same string here?  Or do we just say "If you're looking for
> IVOA-sanctioned DMs, only search by ivoid, the names are somewhat
> arbitrary?"

I favour the last option, deprecating use of Names given that there
is an ID.

Mark

--
Mark Taylor   Astronomical Programmer   Physics, Bristol University, UK
m.b.taylor at bris.ac.uk +44-117-9288776  http://www.star.bris.ac.uk/~mbt/