A few comments from a fly on the wall. Definitely NOT in time order:
Roy Williams says:
> Due to several complaints of avalanche traffic on this mailing list,
> we have tried to restrict discussion among a smaller group of those
> who caused the avalanche yesterday -- and we will include you in future.
Please instantiate this restricted discussion as another mailing list. Folks who want to kibitz should continue to have that opportunity.
He also says:
> Some of us are working on an "alternative representation" of VOtable
> that has the codename V2. The idea is that VOTable and V2 are completely
> equivalent semantically, and can be converted back and forth with no
> loss of content.
I suspect I'm not alone in having the desirability of developing two standards that are completely equivalent escape me :-) Presumably V2 would improve on V1 in issues of efficiency or usability? That being the case, I would suggest that V2 be given its own explicit identity. Don't try to sell it as an evolutionary improvement on V1. Just call it something else and let the market later decide which of the two standards to use. If V1 wins - well, it was obviously good enough. If V2 wins, it was obviously significantly better. Referring to both as "VOTables" will accomplish nothing but confuse the users.
Alasdair Allan says:
> If we keep pushing ahead with the current VOTable stuff we'll reach a
> certain point where we're in lock-in. People will have too much invested
> in the format to start changing things (look at FITS, its awful, but you
> couldn't get rid of it if you tried, and many people have).
I think "lock-in" is a relative term. It is true that the current VOTables are likely already so mature as a concept that they will have to be supported indefinitely. That said, it is by no means obvious that an alternative standard, "VOChairs" or whatever, can't make quick inroads whenever it appears. If it's a better standard that delivers significantly improved services, the major centers and the major VO applications will adopt it gladly.
Regarding FITS and the later discussion of IRAF (bravely defended by Bob Hanisch) - well, folks wouldn't bother to criticize them if they weren't still "best of class". It isn't so much that the community is somehow "locked in" to FITS (other than to support legacy data) - it's that no more preferred format has emerged. FITS and IRAF have many strengths that overwhelm their weaknesses. VOTables will be lucky to have as major an impact on Astronomy.
And he also says:
> Thats why the URI shouldn't point to a file but to something like a
> broker, or an agent, or even a registry, rather than to a file.
Note that there is nothing new under the sun. IRAF OIF images are not only composed of separate metadata and data files in different directories - but the files often also reside on different hosts, and the question arises of whether it is better to access the pixel files via IRAF networking or NFS, for instance. The OIF mapping mechanism isn't as general as URIs, of course, but we did look into the notion of anonymous IRAF networking at one point, with some success :-)
I think the big question facing the various VO interfaces are whether they will be defined *in practice* by written standards (like FITS), or rather *in practice* by software implementations (like - say - Microsoft Word). At some level of complexity, users will have zero confidence in their own ability to implement conforming software. At that point the VO hangs by the thread of its own software. One imagines that a key requirement of the VO remains ease of access. Will that ease be supplied by the elegance of its standards - or rather by the daunting responsibility of maintaining complete and stable crossplatform APIs?
Rob Seaman
NOAO Science Data Systems - and proud son of IRAF
Received on 2004-01-21Z22:39:39