 |
 |
On Wed, 17 Jan 2007, Nathan Vack wrote: > Sure isn't elegant, but as our Real Systems Guys don't want us to > look at the production Oracle instance (performance worries), we've > had pretty good luck screen-scraping holdings and status data, once > we get a Bib ID. Ugly, but functional, and surprisingly fast. We've come at this by setting up a simple web-service that does the screen-scraping. The current implementation is based on ISBNs for our purposes, but could easily be extended to look up via bib-id. After screen-scaping item-status info for a python program, and then needing to do it again for a php script, a Services-Oriented- Architecture article I read 'clicked', followed by an inspiring Access presentation from Richard Ackerman and a BOFeather talk with him and, I think, Stephen Anthony. Since we can't SQL-query our own ILS data directly... (ok, blood pressure is fine again) this solved a lot of issues. A bit of info on the web-service (with some examples), which includes a link to Peter Murray's jester blog article on SOA that prompted this thinking: <http://128.148.7.210/~birkin/wikinotes/doku.php? id=public:soa_josiah_status> As to Michel Doran's insightful post about session-repercussions of http-calls... - I hadn't thought of that, and don't know what if any licensing issues we have regarding that, but I'm curious and will look into this. - We aren't using this service heavily, but if we were, I would think this web-service model would actually scale well in terms of the session issue, because the hits would come from the single web- service 'user' -- as opposed to other models where, for instance, a cron job might initiate a screen-scrape, resulting in each of those instances being perceived as a different user. Does my thinking on this sound right? Last, I set this up not knowing that Z39.50 calls can return status info. I recently found that out by stumbling across a discussion between Godmar Back and index-data folk about php/yaz. <http://lists.indexdata.dk/pipermail/yazlist/2005-December/001453.html> We're now investigating changing the backend to get the data this way instead of the fragile screen-scrape. The beauty of the SOA model is that changing the backend won't have any effect on the systems that call this service. -Birkin --- Birkin James Diana Programmer, Web Services Brown University Library [log in to unmask]