In the early '90s, the sport du jour for the net-savvy was to argue about which search engine was the best. (Some things never change.)

One day, someone decided that it was getting too time consuming to have to check several engines to see which ones were providing relevant results for a particular search term. Many of us remember our frustration with trying Altavista, Excite, Hotbot, Infoseek, Yahoo, Webcrawler and Lycos in succession, trying to figure out which one was doing the best job. What if someone came up with a tool that would check them all, and return a listing of the web pages that scored highest based on an aggregate keyword relevance score?

Thus meta-search was born. One popular meta-search page was called Metacrawler, developed by University of Washington computer scientists Erik Selberg and Oren Etzioni. There were a couple of similar tools floating around, but through a savvy acquisition completed in 1997 by then-small Go2Net, Metacrawler evolved into the category leader. In May, 1999, Etzioni joined Go2Net as its Chief Technology Officer.

Today, there are dozens, and maybe even hundreds, of meta-search tools. Many of them promote themselves by saying "this is cool because it searches all the search engines at once!" And frankly, this often works, because a lot of people haven't heard of a tool like that yet. There is a proliferation of tools with names like SavvySearch, Dogpile, Mamma, Inference Find, C4, Profusion, Findwhat, Searchound, and... well, I'm running out of breath. I see no point in providing a comparison of 5, or 8, or 10 of these, because I might just as well review 200. It's getting out of hand!

In an upcoming instalment, I'm going to check in on how Metacrawler seems to be faring, and give a mention to a couple of other meta-search technologies that seem to be above the crowd.

First, though, now that we seem to be getting past our initial fascination with a search tool that checks all the search engines at once, let's ask ourselves what actually makes for a good meta-search tool. Here are a few criteria, features, and general issues worth considering:

Speed
Today there may be seven or eight, and up to 15, major search engines that are worth checking in a meta-search query. If the code isn't excellent, you'll spend ages waiting for the result.

Consolidation of results
There are a lot of ways to tally up the relevance rankings from the various engines searched. But there are two key ways to report the results of the meta-search query: (1) in serial order or (2) consolidated. Simply listing the search results from different search engines and directories in serial order as, for example, Dogpile does, is highly unsophisticated, and seems to have little "cool value." The best ones consolidate the results. Also useful is a mention of which search engine or engines were responsible for a site's high ranking, and listing the numerical relevance score the site received.

Customization
You may want to experiment with eliminating several of your least favorite engines from a query. For example, many wouldn't want the pay-for-placement search engine GoTo.com (now Overture) included in a meta-search. Some wouldn't want to see directory results from Yahoo, Looksmart, or the Open Directory Project, opting instead for a check of "engines only." For power searchers, the more customization the better. For the average user who just wants to type and go, a modicum of customization is still a good thing.

Ongoing Development
The world of searching the Internet evolves rapidly. There needs to be ongoing attention paid to the design and coding of meta- search tools such that they continue to interact optimally with the underlying search services upon which they piggyback. Here then is possibly the most convincing rationale to use Metacrawler, Proseek, Debriefing, Ixquick, and other meta-searchers which are being improved and monitored by a topnotch developer or development team.

Ingenuity and Innovation
It's easy to get in a rut. Is there a new way of looking at the problem that might be better? As we know, popularity engines, reputation managers, directories, human help, meaning-based search, and many other ideas are being tested in the marketplace by companies looking at navigation from different angles. Why should meta-search be any exception?

Consumer friendly
Why should someone use the service? It is obvious what it does? Will people take to it? Does it provide considerate little touches such as suggested keywords or phrases for "related searches"?

Relevance!
Relevance is a must. It can also be subjective. The only way to truly decide whether a search tool is providing you with the kinds of results that you perceive to be relevant is to give these tools a test drive with some favorite keywords.

No missing pieces
Is more better? Probably. While there is a limit to how many engines you should search at once, a good meta-search engine should leave few stones unturned in its coverage of major search engines and directories.

Boolean Operators and Phrase Searching
A great meta-searcher will treat each search engine correctly to allow your complex queries to be checked properly at each one. This is tougher than it sounds, and few accomplish this fully.

Internationalism
The demand for search tools in more languages or confined to specific country domains is set to explode.

Beyond Generic Web Search
In case you hadn't noticed, search engines aren't the only kind of site or tool which is "going meta". The opportunity to check beyond the web, to newsgroups, music files, audio, pictures, auctions, or pricing on particular goods and services like stereos, autos, or insurance rates, is something that will appeal to many consumers. These other kinds of meta tools are just getting warmed up. Why check out eBay when you can check out all the major auction sites? These are the kinds of questions consumers are asking themselves, and Silicon Valley entrepreneurs are only too ready to supply them with a slew of answers, many of them with little dollar signs attached.