Chad Dickerson emailed me to note a couple of things about Ultraseek. First, the docs. Heh. I stand corrected. REST interfaces are often poorly documented, but not in this case. To wit:
The ql (query level) parameter determines whether the default search form should be simple or advanced.
Here's another cool that Chad points out, which has evidently been added since I first worked with Ultraseek years ago. If you switch from:
http:\//search.infoworld.com/servlet/query.html?col=ifwspid&qt=mycroft
To:
http:\//search.infoworld.com/servlet/saquery.xml?col=ifwspid&qt=mycroft
Then presto! XML results. It's only a short hop from there to an RSS feed that watches for new items matching a search term. Here's an XSLT stylesheet that converts Ultraseek results to RSS 2.0. Here's a prototype feed to which you can actually subscribe. It runs the results of a search for items at InfoWorld.com mentioning 'Udell' through the XSLT transform to produce RSS 2.0.
Granted, that's not too useful as is. The devil is always in the details. In this case, Ultraseek's date ordering (rf=1) isn't yielding all the newest stuff in strict reverse chronological order, which is what you'd want. A real solution would need to collect metadata about actual publication dates. It's just lying around in metatags of IW pages (e.g., <meta name="publicationDate" content="2003-06-13">) and URLs of blog pages (e.g. http://jonudell.net/udell/2003-08-01-infoworldcom-updates.html), waiting to be used.
Still, you gotta love that Web pipelining!
Former URL: http://weblog.infoworld.com/udell/2003/08/01.html#a762