Re: Quick question / idea

[John Rapp <ucsd.jlrapp@edu>]

I think that Infoseek's search engine assigns certain large dimension (256)
vectors to words, then assigns a vector to a page based on the words in it.
Then when you put in a search query, it'll assign it a vector and rank the
ones around it based on how close the vectors are to each other (I don't
know any good web references on this, though). On the special request, I
have no clue, I think the main way search engines find pages is either by
submission, or following links (I think Hotbot does this).

John

At 05:18 PM 1/15/99 -0800, you wrote:
>Has anybody here ever put much thought into -- or done much research on --
how 
>search engines work?  Any good web addresses to look at?  I just had a random
>thought about them for some unknown reason.  Is there a mechanism through 
>which 
>a robot could connect to a known port (possibly re-use the web server's
port?),
>send a specialized request to the server, and get a full directory of what is 
>accessible from that server?  The server would then get to implement on its
>own a way to create a whole directory tree of what URLs it could serve out,
>honoring any robots.txt files that it encounters along the way.
>
>--
>Scott