Re: [IUG] Cover Art using UPC or OCLC numbers


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

I'd like to try and shed a little light on the whole Syndetics cover art based on upc/oclc thing. First, responses to two replies to Lauren's original query; second, some observations based on some testing I did in our staging port. It's a long one; turn back now if you can...


At 08:10 PM 1/13/2009, Catherine wrote:
WNCLN just initiated Syndetics this afternoon, using BIBIMAGE with a url that is
based on ISBN, UPC or OCLC number. So far we have not noticed any of the issues
mentioned. We did just migrate to a new IBM server and will keep our fingers
crossed.

If you're using the BIBIMAGE wwwoption for cover art, you are *not* sending UPC or OCLC numbers to Syndetics, because the current out-of-the-box WebPAC has no mechanism for getting those numbers into the URL. (Check the Syndetics URLs in the source code of a page.) The only way to send UPC and/or OCLC number to Syndetics (or anyone else, AFAIK), is to use WebBridge and the ENH_IMAGE wwwoption (or WebBridge on its own, or some handy dandy scripting).


At 05:54 PM 1/13/2009, Dan wrote:
...UNC Coastal Library Consortium experiences severe slowdown when we enable the ENH_IMAGE web option required for displaying Syndetics cover art based on UPC or OCLC number. I have had a call open about this issue since early Sept. - c1083716 (NCCOA): WebBridge not working with Syndetics Cover Art. When I last checked, the issue had been sent to Software Engineering with no estimated resolution time. We are unable to use this functionality because of this poor response time...

I'm a little confused by the subject of the call---"WebBridge not working with Syndetics Cover Art"--- and why it's even an Innovative call, because it would seem to be a Syndetics problem and not a WebBridge problem. WebBridge's function is to grab the values for UPC and OCLC number from the record and pop them into the Syndetics URL(s). If you enable ENH_IMAGE, configure WebBridge appropriately, do a keyword search, stop the page from loading, then look at the source code, you'll probably see that WebBridge has populated all of the Syndetics URLs with upc and/or oclc values (for those records that have them). In my experience WebBridge does that pretty speedily; the hold up with loading images is not WebBridge (nor the WebPAC server), but Syndetics' response to the calls for images based on the information in the URLs. (If WebBridge is not populating the upc and oclc# values, then it's indeed "WebBridge not working...")


We don't use Syndetics, but I did some testing in our staging port using the Syndetics client codes for three different libraries and got some pretty interesting results. (Note that Syndetics doesn't limit calls for content in any way other than checking for a valid client code.) I configured the Syndetics image URL using one code, did some searches, recorded the time it took to load the results pages, then emptied my browser's cache, altered the Syndetics URL to use the next library's code, then did the searches over again. I played around with sending isbn, upc, and oclc#, and different combinations of these values absent the others. I did it all over again, then all over again a day later.

The most startling thing I observed was that on Day1, it routinely took about 10 times as long to finish loading the same page of images when using Library A's code as it did when Library B's code was used. (Roughly 90+ seconds versus 9 seconds. For a page of 50 results with 40+ images, 9 seconds seemed acceptable to me. 90+ seconds made me want to toss in the towel.) Typical complete load time when using Library C's code was about twice as long as when when using B's. The same server (my WebPAC), the same Syndetics URLs, the same batch of images, but the response time for complete loading of all images varied drastically based on which client code was used in the URL. I don't know how Syndetics does things, so don't know precisely what conclusion to draw from this, but it doesn't seem like correct behavior.

On Day2, Library B and Library C were comparable (and just as speedy as Library B on Day1), but Library A typically took just 2-3 times as long instead of 10 times as long. On Day3, I did real work.

Other observations:

- For books, Syndetics never returned a cover image based on anything other than an ISBN. I.e., if there was no image for the ISBN, having the OCLC number didn't increase the chances of returning a cover image. (This may be consistent with what Syndetics promises for the UPC/OCLC#?)

- For videos, Syndetics returned an image based on the OCLC# about 25% of the time for recent stuff; the percentage decreased with the pub date (to be expected). For the same list of videos, images were returned based on the UPC about 65% of the time for recent stuff. When both UPC and OCLC# were offered, the percentage remained at ~65% with just a few more images returned. I think considering the nature of our video collection---academic library, but with a pretty strong collection of feature films and popular culture stuff---that seems like a pretty reasonable return, but I don't know what (if anything) Syndetics is promising.

- The complete load time didn't vary when images were produced based on UPC (vs. on ISBN).

- A RightResult-sorted keyword search sometimes took longer to fully load than a date-sorted keyword search. Since the top 50 results in a date-sorted results are more likely to have more cover images than the top 50 results in a RR-sorted list (which will usually have old and new titles in the top 50), I think this just suggests that the total load time is faster when Syndetics can find images right off, and they don't have to check multiple values in the image URL before serving up the "empty" image.


Back to real work.


Bob Duncan



~!~!~!~!~!~!~!~!~!~!~!~!~
Robert E. Duncan
Systems Librarian
Editor of IT Communications
Lafayette College
Easton, PA 18042
duncanr at lafayette dot edu
http://library.lafayette.edu/