On Monday 22 September 2003 1:39 pm, Alasdair Allan wrote:
>
> Okay, I'd like to bring up my current pet peeve about the NVO cone search
> standard (ie http://www.us-vo.org/metadata/conesearch/), which is that it
> doesn't include either equinox or epoch.
>
> Both of these are vital for people wanting to do real science with the
> catalogues. If the adopted standard doesn't include these things, people
> doing galactic work won't use the interface because the retrieved
> catalogues aren't going to be valid. We can _not_ assume J2000 for both
> equinox and epoch, it just won't work!
Errm, as I understand it, the business of converting from one equinox to another is straightforward and well understood. So can we assume that the cgi *query* is always J2000, and the catalogue database behind it converts to that using standard libraries?
As for the epoch, as most databases provide the viewing epoch for each object, that should provide enough information? It may create more work at the client end, but it makes it much easier to publish databases, which encourages people to do so.
> I would certainly anticpate a "web service" stage before everyone moves on
> to full OGSA/OGSI grid services, I mean, alot of what we want to do doesn't
> need the complications of OGSI. Why make life difficult for people trying
> to access your service? It would be rather nice if you could talk to a
> service endpoint using straight SOAP as well as a full OGSI compliant
> toolkit if persistence isn't needed for your query...
Agree lots. Publishing data easily, using readily available (whether SOAP and/or cgi-like querying and/or other) techniques, gives us a framework for building all kinds of industry-wide tools, including but not limited to grids.
However I would like to keep an eye on the inherant limitations of using http: it's a simple request/response protocol, with no concept of session state. This makes it simple to use for small queries (which may be 90%? 99%? of likely queries?) but really awkward to use for large queries. State can be added, yes, but it's a layer of extra stuff to add on *top* of http - seems a waste given there's state in TCP/IP connections (*under* http)!
Cheers,
Martin
--
Martin Hill
Astrogrid/AVO, ROE
Tel: 07901 55 24 66
Received on 2003-10-06Z08:43:17