All products featured on WIRED are independently selected by our editors. However, we may receive compensation from retailers and/or from purchases of products through these links.
In debuting its newest operating system Wednesday, Apple will give users a feature that could be reason alone for an upgrade: a built-in metasearch engine capable of searching local drives -- and the Web.
Apple Computer interim CEO Steve Jobs is expected to show off Mac OS 8.5 at the Flint Center in Cupertino, California, said company spokesman Russell Brady. Codenamed Allegro, it is meant to be a bridge between the current Mac OS 8.1 and the upcoming Mac OS X, slated for fall 1999 release.
The new built-in metasearch engine, Sherlock, queries several Web search engines simultaneously and collates the results, an effective replacement for Apple's familiar Find command.
For searching local drives, Sherlock creates a hard-drive index, which is a condensed version of the files. Ryan Meader, publisher of Mac OS Rumors, describes Sherlock as an "amazing technology" that can search the contents of a 9GB hard drive in less than 30 seconds.
Like local indexing, the idea of metasearching on the Web is not new. Client-side search bots, such as FerretSoft's Windows-based search tools, have used similar capabilities for years. But Allegro is the first major operating system to seamlessly incorporate Web and hard-drive searching.
In its early beta versions, Sherlock caused a stir because it bypassed other search-engine portals in returning results. Apple has since negotiated the terms by which Sherlock will use other engines' results.
"Realistically, to sign the big engines on board, they'll have to do some sort of ad passthrough or content passthrough," said Selberg. "The big question is, are they going to be able to make it acceptable to both users and search engines?"
Kurt Losert, acting general manager for Alta Vista, verified that Sherlock would use its search engine, but said the company has come to an agreement with Apple on how results are displayed. Presumably, Apple has similar agreements with all of the search engines Sherlock will use. Apple refused to comment on the issue, however.
Erik Selberg, who co-created MetaCrawler and HuskySearch with Oren Etzioni, said client-side metasearch agents have historically been unsuccessful.
"Basically, people didn't like doing a 'mode shift' in using one application to do a search and another -- the Web browser -- to view the pages," Selberg said.
Bart Richards, general manager of Thunderstone -- a company that has been creating high-performance text-indexing and search tools since 1980 -- added that client-side metasearch is a tricky issue.
The value, he said, is that the servers used by metasearch engines have a lot of bandwidth. They can send a query simultaneously to several search engines and get the results back quickly. Doing the metasearch on a desktop needs connections to the search engines. If the bandwidth is provided by a dialup modem, users should be prepared to wait.
"The biggest value of a metasearch is that you're using the [search engine] server's bandwidth," said Richards. "If you're using a client-side metasearch, you're on a modem -- you've got to make the seven or eight connections yourself to get the answers. There's no free lunch, there's no gain by that."
In contrast, Richards explained, using Web-based metasearch engines such as DogPile or MetaCrawler -- both of which use Thunderstone technology -- give you the advantage of using their bandwidth to connect to other search engines.
But Selberg said that the data transmitted using seven search engines isn't always large. "The big gotcha is if you do anything interesting -- such as download all the URLs each service returned to do analysis -- it can bog down quickly."
Sherlock also has the ability to do natural-language queries. Like Ask Jeeves, search phrases can be typed in plain English instead of in keywords .
From a pure information-retrieval standpoint, it's unclear whether natural-language queries are any better than keywords, Selberg said. Most systems don't do real natural-language queries, he went on to explain; they simply extract "meaningful" terms from a user's query and use those as keywords.
"We've been training people to use keyword lookups since third grade, when they were shown their first encyclopedia," Selberg said. "It's easy, and it works, and people will likely go with that."
Speed is another one of Allegro's big improvements, Meader said.
"Overall, I'd say it's a 30 percent speed improvement. In terms of copying [files], it's closer to 250 percent [faster] -- especially if you copy over the network. Basically, it outruns even NT."
Allegro will be available for purchase on 17 October, with an estimated retail price of US$99.
