On Tue, 19 Jan 1999, Yann Stettler wrote:

> Yes, the filesystem method will always be faster than the HTTP
> one... but with a good server, going through the loopback
> interface, it won't even be that significatif especialy if you
> discard (as the filesystem method do), files that you know shouldn't
> be indexed anyway.

FWIW there's also possible advantage to be had of pulling remote data
through a local HTTP proxy cache, eg Squid (which grew out of the
Harvest web indexing project).

Dan