Re: Knowledge Base (was Re: VI, standards, politics and all that)

[Andrew Wood <ivarch@cus.umist.ac.uk>]

On Tue, 10 Aug 1999, Donovan Rebbechi wrote:

> Might be worth considering query time ( ie we don't want linear time
> searches if we can avoid it ). RH's "search help systems" tool suffers
> from slowness.
> 
> How does glimpse work ? ( pardon my ignoarance ) 

I'm not sure. ;)

The main thing I'm interested in from the guts of `glimpse' is the `agrep'
tool - it's kind of like a "fuzzy" version of `grep', in that it allows for
spelling mistakes and things. Somehow.

Query time is a good point. What I was thinking of doing was searching
through an index of entry titles, and only searching through the actual
entry text if this initial search produced no joy. Maybe.

Don't know how workable that will be, but it can always be changed later.

Glimpse itself uses a special index file to speed up searching in some way.
I've not had a very good look at it yet though.

> Since we make  the distribution, we could just make it part of the
> default. Not a problem.

True enough.

> Easier to use Apache and make the program a CGI script ( this is what I
> thought you were thinking of )

That would mean requiring Apache to be working, though. It seems a bit silly
to have to have a web server on someone's home machine, just to read the
help.

A minimal HTTP server (REALLY minimal, HTTP/0.9 sort of thing) wouldn't be
too hard to incorporate, on the other hand. But there could be problems with
security, I think.

- Andrew