InterOpMay2005 VOTable

---

---

Participants

F. Ochsenbein (FO), G. Lemson (GL), M. Taylor (MT), A. Preite-Matrinez, S. Derriere, R. Hook, J. Lindroos, A. Kembhavi, J.M. Desert (JMD), T. Takata, S. Honda, M. Tanaka, D. Sforna, A. Alan (AA), R. Gutierrez Sanchez, N. Moreau, J. Gupchup (JG), N. Urunkar (NU), B. Hajime, J.C. Mc Dowell (JCM), B. Debray, A. Rots (AR)

Timetable

Time	Content	Leader
14:00-15:00	VOTable applications and Questions
14:00	About C++ Parser and CONVOT (ppt)	Nilesh Urunkar
14:20	About STIL/STILTS/TOPCAT (ppt)	Mark Taylor
14:40	Improvement of the VOTable schema (ppt)	Gerard Lemson
15:00-15:30	UCDs and utypes
15:00	UCD1+ (pdf)	Sebastien Derriere
	utype connection (sxi)	Jonathan McDowell
16:00-17:00	COOSYS vs stc:AstroCooSystem
16:00	STC (ppt)	Arnold Rots
16:15	Discussion
17:00	Other Questions (other datatypes, serializations,...)

Materials and notes from the work group and plenary sessions

VOTable session meeting notes (monday PM)

Introduction (FO)

Starlink VOTable software (MT) (slides in ppt)

STIL: generic table processing library.

• Full implementation of VOTable standard
• Non catastrophic schema changes easily supported
• Table-aware DOM, table-aware SAX

TOPCAT: graphical interface do read/write/update tables

STILTS (STIL Tool Set)

• votcopy : converts DATA encoding
• volint: parsing VOTable and warning for troubles

Discussion

• MT remarks that future developments is uncertain (funding issues)
• Usage : JCM is using it and happy with it -- FO too.
• AA asks precisions on the usage of volint: more intended for VOTable producer than for end user

VOTable C++ parser (NU) (slides in ppt)

• C++ non-streaming parser (supports binary+FITS), using xerces+xalan
• VOTable streaming parser: v2.0 release 2005, april 8 (support VOTable v1.0 and 1.1, and Binary + FITS data)
• VOPlot : visualization toolkit (VO-India) 2D, 3D. Forthcoming release: support multiple table/resources. Integration with tools VOStat, GalICS

Discussion

• MT: How many points in plot?
• No technical limit, depends on memory. Tests are being done, but >1million points with regular desktop machine should work.

• JMD: Ability to plot VOTables with errors?
• Not likely if VOTable is really bad formed. VOPLot relies on SAVOT parser, so it will have same flexibility (true also in case of schema changes: use new version of SAVOT).

• GL: VOPlot as applet: bit hard to use.
• NU: Release of documentation soon.

Suggested changes in schema (GL) (slides in ppt)

No element ref construct

No global elements for attributes and contained classes

Analogy to java binding / inheritance

Discussion

• Not a lot of discussion on mailing list so far
• PARAM inherited from FIELD would be an improvement: agreed
• GL: We could deal with documents only containing TABLE, with no VOTABLE element
• No opposition in the room: suggestion to change schema to version 1.11 with GL's suggestions

ucd and utype attributes (slides in pdf)

Reminder differences between ucd and utype attributes.

Issues: how to distinguish if the contents of the ucd attribute is a UCD1 or a UCD1+? Different possibilities:

• impose that from VOTable version 1.2 only UCD1+ is used
• ucd="UCD1+", and add new attribute for keeping track of UCD1
• ucd="UCD1", and add new attribute for giving of UCD1+

Discussion

Change in schema to accomodate full compliance with UCD1+ syntax (allow ':' char).

Favored solution is to impose the usage of UCD1+ starting with VOTable version 1.2. Better to change number from 1.1 to 1.2 (rather than 1.11) to indicate significant change. We keep the ucd attribute alone, and its content must be a UCD1+

utype syntax (JCM) (slides in ooffice)

• Currently, utype has type "String".
• Use XPATH-like syntax: sed:Segment/Frame/Time/Type
• Write short formal document defining the syntax.

Discussion

MT: Be careful that an exact XPath requirement is maybe not what we want (implies more than we want here).

JCM: ask again about XPath during DM discussion

---

Coffee Break

---

STC presentation (AR) (slides in ppt)

• Need to know: reference frame, time frame, reference position
• Coordinates
• Coordinate Area
• Implementations: STC-X (schemata) and STC-S (string version)
• or link to a template with xi:include (simplify the physical XML documents)
• STC System is extensible

Discussion (see also slides by FO):

• Do we require all time+velocity+space definitions?

• AR shows that Value2ref element allows to point to a column actually containing a vector with the positions, rather that putting the values in Position2D (link from STC to VOTable).
• FO remarks that it would be useful to point from VOTable to STC rather. And also that often the 2 component of the coordinate vector are split in 2 columns: not sure that Value2ref could account for this (SD suggests pointing to GROUP -- need to be explored).

• FO asks cor furrent implementations of STC
• AR indicates that SkyNode and Registry use STC-S

• A VOTable with IDref to STC would validate, but the STC document would not validate because it would not know about the VOTable

• Referencing problem: what if several coordinates in the same system in a single table
• Answer: have different AstroCoord referencing to the same frame is possible

• units specified both in AstroCoord and FIELD : possible contradictions

• FO : how is the Epoch specified ?
• AR : still needs to be detailed

• Conclusion: either STC references VOTable or VOTable references STC. A choice has to be made... before the end of the week

Various points

• MT : suggestion to define an experimental MIME-type for VOTable: application/x-votable+xml before a requesting a stable MIME-type to the IANA: application/votable+xml
• No objections in the room (and suggestion was already made: see this post)

• GL asks if it is possible to reference a row from within another row (like primary key / foreign key)?
• Currently no. Explore what has been done with DM implementations in VOTable

• FO : Add other datatypes to the basic FITS ones currently used in VOTable: date+time, string ?
• Problems: need accurate definition, and issues with binary (for date+time) and FITS (for both date+time and string) serialization
• Conclusion: do not add those new types yet

• Discussions to be held on open topics to reach agreement before the end of the week.

---