Am 01.10.2013 09:06 schrieb James Weinheimer :
> Apologies for cross-posting.
>
> With the US government shutdown, it appears as if all sites with
> .loc.gov (at least) have been shutdown as well. This would seem to mean
> that if someone were using LC linked services id.loc.gov, anything will
> fail.
>
> Interruption with linked data services have been discussed in the
> abstract, but now we have a real example to consider. If someone were
> using the id.loc.gov site extensively for linking name and subjects,
> what would they do? Using other services, such as Amazon, if it goes
> down, you may not get some customer comments for awhile, but that is not
> such a tragedy. Not getting vital information, such as the headings, is
> a completely different matter and would result in failure in the
> catalog. There is the IT concept of "failing gracefully", so that when
> something doesn't work, it doesn't just crash, but is much more
> controlled. http://en.wikipedia.org/wiki/Graceful_exit
>
> I don't even know if anyone is yet using the linked data services from
> LC, but if you were, what would you do now for your patrons? I don't
> know what I would do.
>
> Would it be possible to "fail gracefully"? This "downtime" could serve
> as a time of reflection into some of the realities of linked data that
> we all know can and will happen (again).
If you had not only linked data, but linked "open" data, you are allowed to
build up your own service with the data. Thus you have mirrors. For us, we have
such an internal mirror of e.g. the German authority file GND and the German
ZDB Isil authority file . Thus, in principle, there is no "single point of
failure"[1].
Now, your are right, not every single small institution can build mirrors. We
provide also a public SPARQL endpoint where you can lookup the GND URIs, so
there is no need for everyone to build mirrors on their own.
What is needed then would be a list of server where these redundant, mirrored
services are hosted. Then, because the URIs themselves won't work, as the
domain of the URI is down, your service had to use some logik to switch to that
other server denoted on the list.
-o
[1]https://en.wikipedia.org/wiki/Single_point_of_failure
Received on Tue Oct 01 2013 - 04:43:57 EDT