where next?

[Clive Page]

I'm taking up the challenge to get involved with this again, but just to
question one recent point:

On Mon, 16 Jun 2003, Robert Hanisch wrote:

> My point about "small enough in number" is that if the number of elements is
> too large, resource publishers will either 1) not bother to fill them in, 2)
> fill them in with rubbish, just to get on with things, 3) not bother to
> register at all.

I may well have missed some relevant postings, but this makes it look as
if those responsible for each data center/service have to fill in the
relevant registry details by hand.  If that's really true, then we ought
indeed to make things as simple as possible (or perhaps a bit simpler than
that).

I had hoped, however, that we could set up a structure of Web Services
which allowed some central registry-robot to discover the gory details
themselves.  If each service publishes its interfaces using WSDL, and
there's a suitable structure to them, then this putative robot should only
need to be given the top-level URL at each site, and can then go away and
discover the rest. For example, for each service the robot find all the
databases, and for each database it finds all the tables, then for each
table it finds all the column names and other tabular metadata, and so on
down until it reaches the end (or some level not of interest).
Registries with different level of detail can be set up simply by setting
up the relevant robot with a different setting for its maximum delving
depth.

If this principle IS accepted, than I think our design of the registry
ought to impose constraints on our Web Services and their WSDL interfaces,
so that they will allow robots to fill the registry in this way.  In that
case, we ought not just to design the registry but also design and build
the robot.  And those designing VO Web Services ought to know that they
need to design in robot-searchability to their metadata.

If we do NOT set up a structure which permits this, I think the management
overhead will become overwhelming, and resource descriptions will never be
adequately up-to-date, because many of our data archives are under active
development, and their contents always changing, but their managers will
never have enough time to update their own registry entries.  Experience
of the Internet shows that hand-compiled directories simply don't scale,
but that robot-generated lists can produce adequate search engines.


-- 
Clive Page,
Dept of Physics & Astronomy,
University of Leicester,    Tel +44 116 252 3551
Leicester, LE1 7RH,  U.K.   Fax +44 116 252 3311