Problems aplenty getting to grips with the search facilities provided by Google's Mini appliance.
I’ve spent much of this week struggling with a Google Mini, a server that plugs into a network, crawls what it’s told to crawl and then returns search results using the vendor’s famous technology.
The Mini had been sitting there in perfect anonymity – as far as being bright yellow allows – on the server rack, looking like a slab of emmental. It’s a cheaper and therefore more limited version of the Enterprise search which I first came across over five years ago while working for Research Machines.
One fine day, someone vaguely remembers the Mini’s existence and I’m asked to “look into it”. Good things rarely come of these three words.
I’ve never been keen on Google’s documentation and the bumph that ships with the box is no exception.
The information is properly formatted but it just never seems to get inside my thick skull. In my experience as an information designer, this tends to happen when the writer fails to provide a context. It helps to know the purposes, reasons and consequences of one’s interactions.
I’m not a sys-admin, but I’m not exactly a technology virgin either. Even so, getting to grips with the new vocabulary (“OneBox”, “Host Load Schedule”, “Freshness Tuning”) is a difficult experience.
According to the log file, I learn that the Mini has been happily crawling away for the last almost three years already. Perhaps the licence limit has been met.
Could this go some way to explaining the steampowered search speeds I’ve been getting? With 50,000 items of content in the index and me the only user querying them, an enormous recordset of 269 results is rendered in a dizzying 34 seconds!
One of the better aspects of the box admin is the ability to alter the search/results UI. There’s so much bad code in the factory XSLT that this comes as a big relief.
Customised search results
As this autumnal week of long shadows draws to a weary close, I’m left plugging away at what originally promised to be the most interesting aspect of the Mini for me.
OneBox modules are self-written keys to leverage information from other datasources when returning Mini search results. In theory, you can poll another server when users ‘trigger’ the functionality in their search terms. Here’s an example:
- User submits a search query for who: mike padgett
- Google Mini checks the query for triggers – keywords or phrases that trigger a OneBox
- Google Mini runs the search for mike padgett normally and polls the location defined in the OneBox, say, http://myserver/?q=mike+padgett
- Google Mini processes the results and returns custom-formatted content above the ordinary search results
This offers fantastic opportunities for the “joined-up” organisation. Rather than simply go ahead and dump the output of the crawl matches, we can first offer customised boxes of information. Google already uses many examples of OneBox modules in its own online searches, including geographically aware weather summaries and unit conversion.
I decided to adapt an “internal organisation” example outlined by Google by developing a little box for staff member contact details, triggered by a who keyword similar to that mentioned above.
So far, however, I’ve been unable to get anything working. What I’ve been left with are: a blank search log, no error logging and a mild headache. Watch this (vacant) space…
It’s not easy to say whether the Mini – or indeed big brother Enterprise – has been a success. Of course, the Gartners of this world probably told you that it was the best thing since Sharepoint.
As for the people who actually have to work with the stuff on which management bursts the budget, there are some pretty jaded reviews out there.
Finally, I think that Ville Laurikari sums up many of my own frustrations when he says: “It’s like shining a bright light on a big pile of crap. It’s still a pile of crap, but you can see it more clearly.”