[VEP-003] datalink/core#sibling: Ready for TCG

[Markus Demleitner]

Dear DAL,

On Tue, Feb 18, 2020 at 11:20:53AM +0100, Laurent MICHEL wrote:
> I agree we need a very generic term to say that this dataset is derievd from
> that one.
> Taking the place of a lambda-user, I'm pretty sure that neither #sibling nor
> #crosseddataset make sense.
> 
>     #sibling is too vague
>     #cross... reminds some crossmatch process
> 
> In contrast, #see-also is something everyone can understand.
> The meaning of the "also" suffix can be resolved by a subtype in
> content_type
> 
> So my vote goes for #see-also

While I'd deny that #sibling is vague (it's very well-defined, I'd
say, much more precise than #see-also), I give you I have some
nagging doubts whether it's *useful*, in the sense of "it expresses
something actual clients are interested in".

It all comes down to finding a good compromise between the needs of
service operators -- I think #sibling does an excellent job there --
and the needs of clients (and those might rightfully ask: Well, how
do I present this "sibling" thing to my user now?).

On the other hand, when I try to come up with a definition of
#see-also, I'm cringing a bit, too.  Actually, I've tried for five
minutes, and nothing even started to work out.

So:

(a) if you think #sibling is a good and useful idea (preferably with
ideas on what we say clients ought to do with them): Please speak up
now.  As things are, it could well be that VEP-003 will die.

(b) if we go for #see-also, we'd need a VEP-004, and for that, in
particular, a description that says what it means -- a description
precise enough so service operators now when to use it, and good
enough so client writers know what to do with the links annotated
with it.

Having said that, I think it's still a good idea to keep the big use
case we have in mind: Gaia time series and spectra.  I suspect if we
have a good mock-up on how we think clients should present these to
the user, that might help a lot to come up with a concept that's
useful and precise.

       -- Markus