Call for contributions: Major version transition

[Markus Demleitner]

Dear Colleagues,

[apologies for the wide cross-post; followups to registry only, I'd
suggest]

As the VO matures, we will at some point have to introduce changes in
our standards that break previous, widely-deployed standards.  For
instance, some of the proposals put forward by the Protocol Transition
Tiger Team would, if adopted, render existing clients broken; much more
mundane, Simple Cone Search, as one of the first standards defined by
the IVOA, now really shows its age.  Bringing it up to date will again
be a breaking change and hence, in our language, a major version
transition.

Our first real attempt at a major version transition, SIAP, so far
has not been a resounding success; almost 10 years after SIAv2 has
been released (on 23 December 2015), there are still many SIAv1
services coming online[1], and there is no telling how much data you
would miss if you did a SIAv2-only query.

We have therefore started a document on how to do major version
transitions in the future:
<https://github.com/ivoa/major-version-transition>: this may eventually
become an IVOA Note, but for now we would like to have the community's
opinions on what we have written -- and also gather experiences with
successful, failed or so-so major version transitions *in federated
systems*, quite possibly from beyond astronomy.

To that end, we plan to hold a session at the upcoming College Park
Interop.  The rough plan is to have some short-ish talks introducing the
matter (and the troubles) and then at least 45 minutes of discussion
time.

If you have

* comments about or suggestions for the major version transition
  document, or
* stories of non-trivial (i.e., >=3 implementors, >=20 independent
	deployments) major version transitions and their outcomes,

we would like to invite you to give a short talk at this session
(remote talks are of course ok).  For instance, if you remember the
UCD1 -> UCD1+ process, or want to moan about the adoption of ADQL2.1
(admittedly, that's a minor version transition, but still), we'd like
to hear from you.  Or if you have hard data on how much we would lose
if we officially declared SIAv1 as deprecated and told people to
ignore the respective services, or have non-VO experiences where
someone got a whole community to cooperate on sunsetting an outdated
standard: Please send a mail to me with a title and a short (a few
sentences max) abstract.

Thanks,

          Markus (for an anonymous ad-hoc working group that you'd be
            welcome to join)

[1] Don't take my word for it.  For a list of SIAv1 resources that were
registered in the past 12 months, run

select ivoid, created
from rr.capability
  natural join rr.resource
where
  standard_id='ivo://ivoa.net/std/sia'
  and created>'2024-05-01'

on the RegTAP service of your choice [shameless plug: if you don't have
one, try https://dc.g-vo.org/tap]