IVOA-science in Softid

[Markus Demleitner]

Dear Colleagues,

As a small piece in the puzzle of how to deal with stupid crawlers
overloading our services, I have occasionally suggested a small
secret handshake between legitimate clients and our servers in hopes
that might be enough to avoid the APIkeycopalypse.

I am, indeed, rather convinced that "follow a standard" would already
work pretty well to keep out most "AI crawlers", and that
quasi-standard could be our note on operational identification
of software components <http://ivoa.net/documents/Notes/softid/>
(which has quite a few other use cases, too, so if you are writing VO
clients or servers, please have a look).

For clients, this note has the provision of an IVOA comment.  So far,
the idea has been that validators would have something like

  User-Agent: STILTS/3.1-4 (IVOA-test http://validators.org/results) Java/1.8.0_181

in their request headers (crawlers would give IVOA-copy), and
diligent beancounters would remove such requests (done for validation
in this case) from their usage analyses (as questionable as these may
be even then).

So far, the standard said that "normal" usage should not give any
purpose.  Now, giving "science" as a purpose would exactly be such a
follow-the-standards handshake.

I have now written up a proposal for how this would look like as
a PR against softid: <https://github.com/ivoa/softid/pull/1>.

What do people think?  As a client, would you adopt it?  As a server,
can you imagine any use for IVOA-science in user agents?

If sufficiently many people feel this is a good idea: Perhaps it's
then time to try and get softid endorsed?

Thanks[*],

         Markus

[*] Enjoy your holidays if you have them!