SIAP2

[Doug Tody]

Hi Roy -

> I'm not too sure if there is an exploder to send this to -- could you please
> make sure that my suggestions (below) goes to the right group who are
> specifying SIAP2?

This would be the DAL exploder.  I will copy this response there.  Some
yet to be specified subgroup of the IVOA DAL working group will specify
the next version of SIA.


> (1) Replicas
> In a SIAP service that I have built, the image dataset is available in
> multiple places and protocols. Therefore I have added a new keyword
> "dataserver". If this keyword is set to X, then it controls the nature of
> the URLs in the resulting table of images. In my case, I have values
> understood for attribute:
> -- if it is "http-cacr", then the URLs point to an address at
> cacr.caltech.edu with the form http://
> -- if it is "http-npaci", then the URLs point to an address at nvo.npaci.edu
> with the form http://
> -- if it is "srb-npaci", then the URLs point to an address with the form
> srb://

If I understand this correctly, a default query will output references
for all of the above, with the "dataserver" attribute set for each image
reference in the output votable.  If one then sets "dataserver=XX" in the
input query only images served by the named server/protocol will be listed.
Is this correct?

I guess the reason for this is to allow the server location to be understood
and specified without having to parse the URL.  We probably also need
the collection ID and dataset ID (which would be the same for all replicas)
to make this scheme work.

 
> (2) Logical names
> When images are downloaded to a local disk, I would like to know what to
> call the files. Currently my only option is file1.fits, file2.fits, ....,
> because there is only a URL, no logical file name (LFN). Some URL's are
> constructed as camcol=3&rerun=6 etc etc and not easy to convert to a
> reasonable file name. Some have .fit.gz and other suffixes that should not
> be part of the LFN. Thus if I want to run the same computation again, I do
> not know what files are already in the local cache. It would be really nice
> for the table of images to include a column for the logical name -- not
> necessarily mandatory, but with a specified VOX:UCD for it. "If I already
> have a file of this name, then I don't need to download the URL".

The Web-way of doing this is to cache things based on the URL string.
You wouldn't use this as a filename, but in some sort of index that points
to the filename.  This is the concept for the current SIA and is part
of the reason we have a static, fully specified URL acref.

There would be no harm in providing a suggested filename in the output
metadata (FITS writers commonly do this for example).  I'm not sure how much
this would accomplish here though, since we are not pretty-printing file
names for the user, and your client code would already have to deal with
services that don't include this optional attribute.

The default filename could also be based on the collection ID and dataset
ID, if we had those keywords.

	- Doug