The problem with Web searches is that they are Web searches — humble explorers can explore for key terms or concepts, but sorting and limiting by associated metadata is all-but-hopeless, and even the special terms people can use to narrow a search are not especially useful.

I was frustrated this morning whilst seeking information about battery-powered generators, to support an article I’m writing on emergency generation systems. The challenge is that, thanks to the awesome power of SEM/SEO, commercial sites hit the top of the site rankings charts. Useful sites (that is, non-commercial ones that aren’t blacklisted by my contract client), get lost in a sea of sales materials and poorly written articles intended to drive up AdSense revenue.

Current desktop-search features allow us to find files or substrings after limiting our selection criteria in dozens of possible ways. I can find a file that was modified within the last three weeks and is less than 25K in size and which was sent to me as an email attachment — all with no problem.

So why not something similar with the Web?

Yes, I know that the Web does not have handy sort points. There is no universally accepted META tag for “I’m a business site pretending to provide useful information so you will click through to me and maybe buy my product.” I get that. But with the powerful algorithms governing search nowadays, it seems that the major search providers (Google, Yahoo, Bing) could do something by way of multi-level/metadata searching.  I’ll gladly settle for 80 percent of a loaf.

Here’s a thought: Harness the oomph of Google and Microsoft to force a revision of the HTTP protocol to include a handful of volunary, basic, standard, indexed search parameters that get embedded in the META tags. To reduce manipulation, searchers could vote on how well a particular site conformed to the expected results of the search, with low rankings subject to downranking in the search engine, or even outright blacklisting.

In short, I want to partially bypass search for author-supplied tags or keywords and instead refine my search according to characteristics about the authors themselves.

The first engine that can crack metasearch, has (in this humble author) a dedicated user for life.