[Datalink] and RESTful services

[Patrick Dowler]

This was an aside to my previous post, but I thought it best to separate 
(TL;DR and all that).

As an aside, for RESTful web services it is true that service descriptor 
cannot currently describe how to call them. It can tell the users about 
the service via the standardID, but this is considerable effort (maybe 
by GWS?) to figure out how to describe RESTful services. In general, I 
don't see a generic REST client being able to do something sane without 
grok'ing the semantics of the various resources so IMO we need to not 
worry about REST in DataLink-1.0 and

Even then, if you make a descriptor for the vospace {nodes} capability, 
you cannot really include a reference to a FIELD with "vos" URIs to use 
with it... it is maybe a bad example because with vos URIs the client 
can in fact extract the resourceIdentifier so it could find the right 
service descriptor that way (vs lookup in registry).

In our vospace impl, we have a custom resource with fast parameterised 
transfer negotiation: that can be described fully and a generic client 
could successfully get a file using it... this is in the vospace 2.1 WD.


I expect the first real case that will come up in DataLink is trying to 
describe an AccessData type of service that is async and using UWS for 
job creation and management. We'll need some way to say that the 
capability is async. The standardID essentially requires clients to 
"know" the service spec and doesn't currently cover custom servcies... 
or can it?

- for a standard service, just specify the normal standardID and expect 
the client to know if that is UWS or not

- for a custom sync service, there is no standardID

- for a custom async (UWS) service, set standardID to be the UWS 
standard (generic async job)

In all cases, the input parameters are described. Is that a total 
perversion of standardID? Would it be wrong to do the same in a 
VOSI-capabilities record for a custom async service? I think the two 
should carry the same info so the latter is the real question...



-- 

Patrick Dowler
Canadian Astronomy Data Centre
National Research Council Canada
5071 West Saanich Road
Victoria, BC V9E 2E7

250-363-0044 (office) 250-363-0045 (fax)