>As for search engines, those can be hooked in independently; which some
>have done with mhonarc. It is a waste of my time, and probably other
>developers of mail processors, to write search engines when one can
>already utilize well developed ones like Lycos, Glimpse, etc.
I've been messing around periodically with a HyperMail-like gizmo for a
couple of years now. (You could see its output at http://www.mccmedia.com/
if the server weren't dead, behind a locked door right now, but it should
be back up tomorrow at the latest. This is the same machine known also as
asearch.mccmedia.com, where I've been doing about 50 Web-related lists.
It's been a mess lately, because I'm doing the conversion on a machine at
home and ftp'ing the files to the server at Verity, while trying to get
ISDN working... Say, y'all don't care, do ya?)
Anyway, ISDN is working (at last) and I've reached a deal with O'Reilly &
Associates to do a series of books on Internet scripting (JavaScript,
AppleScript and VB are planned), so I'll surely re-implement and improve my
gizmos with Java/JavaScript. (Examples from the book & all that.)
The suggestions here are great.
>> 4. Allow relational queries: by date, author, subject, message-id,
>> keywords, or any combination. Essentially, treat the archive as a
>> relational database table with fields message-id, from, date, subject,
>> keywords, and body.
>
>This is best done by utilizing an existing database system (eg Oracle),
>and using mhonarc (or other prefered mail->html filter) to convert
>retrievied messages to html on-the-fly.
Our search engine allows relational queries as well as full-text queries,
so I've recently prototyped a version that will generate the menus and
threads on the fly, as searches. Thus, the user can choose a starting
date, number of days and/or messages to see, as well as applying other
filters. Those filters can be saved as "topics," in our lingo, which means
that they'll be pre-indexed as new documents are added, so that search
results are quite zippy. (See the Verity marketing mumbles at
www.verity.com for supporting arguments.)
I'm quite interested in adding a Harvest interface, as well as implementing
the sort of rating system that was mentioned here. In fact, I think that's
the most interesting potential improvements... other involve collaborative
knowledgebases, "scout" robots that look for related Web resources... many
ideas, only so many hours in the day. But now that I'm not confining
myself to languages that run on computers design in Cupertino, perhaps some
collaboration is in order.
In particular, who else is interested in working up some of this stuff in
Java? I need those examples for the books... ;-)
Nick