Initial survey of VO standards – which do we need/want to adopt?
Discussion items
VOTable
SUI/T is concerned about efficiency of data transfer with VOTable for large results.
VO BINARY2 may be a reasonably performant alternative. Need to check if reference implementations of ser/des are available for Java, C++, Python.
Could we work with IVOA to get an additional encoding officially adopted, based on more contemporary and widely adopted ser/des frameworks (e.g. protobufs or capnproto)?
These frameworks have advantages (and disadvantages) of automatic stub generation for multiple languages.
Could be generic row-based result protocol; would not have to be touched/changed for every different result row type. This is done currently for Qserv, where protobufs are used for the wire ser/des for query results flowing from workers to czar.
It would be helpful to have a working example to demonstrate efficiency wins to IVOA in order to advocate adoption. Consider a test based on existing Qserv protocol description.
An advantage of getting IVOA adoption is that we could then perhaps dispense with poorly performing encodings which are otherwise currently mandated by IVOA.
Tim notes that science pipeline is currently looking at migrating afwtable to astropy; once this is done it is much easier to go from astropy to VOTable.
Firefly does have a VOTable parser.
TAP
IPAC has some very preliminary experience accessing other catalogs via TAP. IRSA has just published via TAP.
IPAC has work on their plate to adapt Firefly to TAP, but hasn't begun this work yet beyond a very simple prototype.
IPAC considering/recommending working with Mark Taylor (TOPCAT) in consulting capacity.
dbserv will be adapted to TAP; Brian will be undertaking this work.
Brian has some preliminary concern about slightly different semantics of sync/async between TAP and Qserv as currently formulated. We shall see...
Need to coordinate this work so we don't break/stall SUI/T pending TAP client-side work.
ADQL
IPAC has a simple ADQL parser in C++. Qserv team should check this out and see if we could leverage this for upcoming parser work. Currently targets Oracle, but we will look anyway
SIA
Simple cone search, etc. may overlap with imageserv/metaserv. Should we adapt our services to SIA, include SIA as alternate access method in our services, or advocate the IVOA adopt our protocols here?
Tim recommends we also take a look at VO data-link service (as an alternative to SIA hardcoded approach?) for drilling down from catalog results to images.
VOSpace
SUI/T will be driving an investigation of this, with respect to L3 user workspaces, etc.
There may be some involvement from dbserv though (for example, request that sends query results to L3 workspace rather than streaming results directly to client).
CAOM
MAST and IRSA are considering moving to CAOM
Adoption could affect some of our table designs
Would need to coordinate this with data production and alert production for L1
Maybe something new ("obsserv", "visserv"?)
VOSI
Defines minimal core of VO which must/should be implemented. May be a good place for us to look first.
Mandates some VO registry things.
MAST provides a registry; could we publish into theirs?
IPAC/IRSA experience with MAST registry was not positive.
Tim mentions that JWST is motivating cleanup/maintenance here, so may be improved/improving.
Operational def'n for sucess with registry: can TOPCAT find you?
Misc
How do we pick VO standards that are well-adopted "winners", vs. new fads that will be dead in a couple years?
Could MAST tell us user statistics?
Brian might have some folks at Goddard to talk to.
XiuQin's experience is that TAP, SIA, SSA are de-facto expected.
Bonus lengthy digression into remote-Butler design and issues, a la mode...
Let's have a follow-up meeting at a time when we can get both Gregory and Jacek involved.