Re: WWW meta indexes (proposal)

joe@mit.edu
Wed, 27 Oct 93 15:20:47 -0400


For those who have asked about technical details regarding the GNA
meta-library.

I'm using the Postgres object-oriented database (it's a successor to
Ingres). The search language is something called postquel and there
is a perl-interface to the database. I use the plexus server with
some additions to convert the WWW address to postquel commands and to
parse the response.

The big win with postgres is that searches are relatively quick and
allows some very sophisticated search terms. Also it would be trivial
to write parsers that read in or out things into the database.

The minuses are that the record lengths are limited to 2048 bytes and
the server is somewhat buggy. Also postquel is not a standard
language.

As for portablity, I think my system will run on any system that runs
postgres (which includes Sun's and Decstations). I don't have any
objection to giving away the code and database as long as Global
Network Academy is giving credit for putting together the system.

As for what needs to be done..........

1) A spec needs to be written describing the fields in the database.

2) A mail robot needs to be written that parses the input format. The
easiest way of implementing this to have a perl script that reads the
incoming file and stuffs the fields in an associative array. It's
then trivial to take the associative array and stuff it into the
database. Also it would be a good idea to write a procedure that
takes an associative array and converts it to index format.

It's crucial to have a mail robot handle requests semi-automatically,
because I'm overloaded right now trying to add requests by hand.

3) Some scripts need to be written to allow access of the database for
people who do not know postquel.