Paul Walk writes
> With the current ubiquitous harvesting protocol (OAI-PMH) I think
> that this approach, although logical, is just not viable. It is
> non-trivial to maintain even a single layer of aggregation with this
> protocol as the state of the resources being aggregated is not
> reliably conveyed and the fidelity of the aggregation starts to
> drift over time.
This what I felt when I took part in the design of OAI-PHM. I
pleaded for a dumb file based approach that could be transported
with various transports protocols. http, ftp, rsync. Nobody took me
serious.
> This is the kind of problem which has led Herbert Van de Sompel (and
> others) to begin work on a better protocol, ResourceSync. However,
> it will be some time before this is a wide-spread as OAI-PMH,
> assuming it even gets wide-spread adoption.
You can have a as many geeks divising as many transport protocols as
you like. This is not where the problem with the current
infrastructure is. The problem is that repositories have all sorts
of digital stuff in them, with no way to sort out what would be
of interest for further processing. This makes building value
interesting added services a costly proposition.
--
Cheers,
Thomas Krichel http://openlib.org/home/krichel
skype:thomaskrichel
|