msdemlei
commented
4 years ago On Mon, Feb 10, 2020 at 12:26:07PM -0800, Patrick Dowler wrote:
pdowler commented on this pull request.
- -\subsection{Capabilities: VOSI-capabilities}
-A standalone DataLink web service must have a -VOSI-capabilities endpoint as defined by DALI and VOSI. -The standardID for the \blinks\ endpoint is +Operators still wishing to declare \blinks~endpoints can do this by
Do we want to mention including a links capability in a data collection resource -- to make publisher_did values resolvable? We have prototyped that and made it work internally (I have not yet modified our published registry resource types/content for this)... we use it to find several endpoints that accept publisher_did value as the input, one of which is {links}.
While I agree that in principle, using datalink for PubDID resolution might be a good idea, it's possible in several places in the GAVO datacenter, and it's even supported in the global PubDID resolver (http://dc.g-vo.org/ivoidval/q/didresolve/form), I'd say that some other document would first have to make a stronger case for resolving PubDIDs in the first place (IVOA Identifiers comes to mind; after all, it's where the pubDID form is recommended, http://www.ivoa.net/documents/IVOAIdentifiers/20160523/REC-Identifiers-2.0.html#tth_sEc4.1).
Another reason why I'd be reluctant is that when describing our ID parameter we say
The ID values are found in data discovery services and may be readable URIs or opaque strings. If an ID value specified by the client is not understandable by the
-- which doesn't require (resolvable, globally unique) pubDIDs, and I agree with not requiring it.
Hence, if we said anything about global discovery in the capability section, we'd have to say "works only for those of your IDs that actually are ivoids that have base ivoids pointing to records with datalink capabilities".
That's all not terribly difficult, but it's a noticible effort, and that I'd rather avoid for something that nobody so far has been asking for.
This PR tries to remove material that, based on what I wrote in the caproles note, we will probably regret later, and that isn't actually used in datalink operations anyway. Positive side effect: shorter document!