unAPI usage

[Binkley, Peter]

I was excited the other day to notice that we've had a big surge in use of our unAPI service attached to our opac (http://www.library.ualberta.ca): since September we've seen 1300-2300 MODS records downloaded per month, up from a couple hundred/month. I assume this is because of the spread of Zotero. At first I thought these were records actually saved by users, and got excited by this level of usage. A little testing, however, showed that Zotero retrieves the record whenever it finds a unAPI link (presumably to determine the item type so it can show the right icon in the location bar). So there's no way to tell from the server logs whether the user saved the record or not.

Just curious: are other unAPI implementers seeing a big increase in downloads this academic year?

Peter

[Robert Forkel]

not really:
grepping our access logs for the last 5 months gives:

robert@srvd:/var/webstat/logs/resolved-srvp$ gunzip -c
lrr/access-08-01-* | grep "/unapi?" | wc -l
220752
robert@srvd:/var/webstat/logs/resolved-srvp$ gunzip -c
lrr/access-07-12-* | grep "/unapi?" | wc -l
319427
robert@srvd:/var/webstat/logs/resolved-srvp$ gunzip -c
lrr/access-07-11-* | grep "/unapi?" | wc -l
189454
robert@srvd:/var/webstat/logs/resolved-srvp$ gunzip -c
lrr/access-07-10-* | grep "/unapi?" | wc -l
213381
robert@srvd:/var/webstat/logs/resolved-srvp$ gunzip -c
lrr/access-07-09-* | grep "/unapi?" | wc -l
269213

that's for relativity.livingreviews.org
we offer unapi for our own publications' metadata and for our
reference database.

[Robert Forkel]

i should note that the numbers are this high, because we also link to
the unapi records, so they are indexed by search engines, too.

[Binkley, Peter]

Could you maybe filter out the directly-linked hits by looking for the
initial unAPI calls for the menu of format types? Zotero and presumably
other unAPI clients produce this pattern:

xxx.xxx.xxx.xxx - - [30/Jan/2008:11:51:13 -0700] "GET
/unapi/server?id=3295012 HTTP/1.1" 300 318
xxx.xxx.xxx.xxx - - [30/Jan/2008:11:51:15 -0700] "GET
/unapi/server?id=3295012&format=mods HTTP/1.1" 200 4726

So "| grep -v format" might do the trick.

Peter

[Robert Forkel]

good idea (sad results):

robert@srvd:/var/webstat/logs/resolved-srvp$ gunzip -c
lrr/access-08-01-* | grep "/unapi?" | grep -v "format=" | wc -l
9
robert@srvd:/var/webstat/logs/resolved-srvp$ gunzip -c
lrr/access-07-12-* | grep "/unapi?" | grep -v "format=" | wc -l
14
robert@srvd:/var/webstat/logs/resolved-srvp$ gunzip -c
lrr/access-07-11-* | grep "/unapi?" | grep -v "format=" | wc -l
32
robert@srvd:/var/webstat/logs/resolved-srvp$ gunzip -c
lrr/access-07-10-* | grep "/unapi?" | grep -v "format=" | wc -l
22
robert@srvd:/var/webstat/logs/resolved-srvp$ gunzip -c
lrr/access-07-09-* | grep "/unapi?" | grep -v "format=" | wc -l
9

[Alf Eaton]

Binkley, Peter wrote:
> A little testing, however, showed that Zotero retrieves the record whenever it finds a unAPI link (presumably to determine the item type so it can show the right icon in the location bar).

Yes, and it seems to do this in preference to using embedded metadata like COinS, too, which is a couple of possibly unnecessary requests for every page view.

alf

[Binkley, Peter]

Wow - so is the spread of Zotero a local thing? Or is Zotero getting
your metadata by some other means than unAPI? Or is my traffic from
some other source? (Maybe I'll turn on logging the user agent and see if
that tells me anything).

[Robert Forkel]

On Jan 31, 2008 11:42 PM, Binkley, Peter <Peter....@ualberta.ca> wrote:
>
> Wow - so is the spread of Zotero a local thing? Or is Zotero getting
> your metadata by some other means than unAPI? Or is my traffic from
> some other source? (Maybe I'll turn on logging the user agent and see if
> that tells me anything).

although we also offer coins, zotero does use the unapi data. i'd also
guess that our readers are not exactly the avantgarde of zotero users.
but i don't really know who that would be anyway. fact ist we are not
tied to a university, so effects of student training courses would not
show, for example. an additional explanation of the low numbers is
that our readers mainly come for the fulltext articles, less so for
the bibliographical records of references cited therein. so basically,
i guess we our data is not really comparable.