[Ops] Inconsistency between full searchable registries

Markus Demleitner msdemlei at ari.uni-heidelberg.de
Tue Mar 15 10:33:36 CET 2016


Dear Registry and Ops,

On Tue, Mar 15, 2016 at 09:02:58AM +0100, Menelaus Perdikeas wrote:
> I think that race conditions are also possible in the design of the
> OAI-PMH protocol itself (but these can be addressed for practical
> purposes by using comfortable overlaps in incremental harvests). 

Just as a reminder what this race condition is (mostly relevant to
the implementors of publishing registries, and it'd be great if you
could avoid it in your implementation or usage practices) -- in

http://ads.ari.uni-heidelberg.de/abs/2014A%26C.....7..101D

in section 5 (p. 4 right column in the arXiv preprint) there's an
explanation of that race condition.  

TL;DR: Only use the resources' "updated" timestamps as OAI-PHM change
date if OAI-PMH sees the records in the instant of the update (and
that's independent of the VO, that's true for all applications of
OAI-PMH).


Cheers,

       Markus


More information about the ops mailing list