Received: from mailin2.email.bigpond.com (juicer14.bigpond.com [139.134.6.23]) by postgresql.org (8.11.3/8.11.1) with ESMTP id f6NH3nf61924 for ; Mon, 23 Jul 2001 13:03:50 -0400 (EDT) (envelope-from giles@nemeton.com.au) Received: from nemeton.com.au ([144.135.24.72]) by mailin2.email.bigpond.com (Netscape Messaging Server 4.15) with SMTP id GGXRO000.AX9 for ; Tue, 24 Jul 2001 03:09:36 +1000 Received: from CTPP-p-144-134-40-25.prem.tmns.net.au ([144.134.40.25]) by bwmam02.mailsvc.email.bigpond.com(MailRouter V2.9g 8317/3467803); 24 Jul 2001 03:09:36 Received: (qmail 12517 invoked from network); 23 Jul 2001 17:03:34 -0000 Received: from nemeton.com.au (203.8.3.33) by nemeton.com.au with SMTP; 23 Jul 2001 17:03:34 -0000 To: "Josh Berkus" cc: pgsql-sql@postgresql.org Subject: Re: Full-text Indexing and Primary Keys In-Reply-To: Message from "Josh Berkus" of "Mon, 23 Jul 2001 08:32:45 MST." Date: Tue, 24 Jul 2001 03:03:34 +1000 Message-ID: <12515.995907814@nemeton.com.au> From: Giles Lean X-Archive-Number: 200107/313 X-Sequence-Number: 4121 Josh Berkus wrote: > 1. Can anyone explain to me what "full-text indexing" is, and why we do > or don't need it for Postgres? The marketing types keep asking me about > it ("buzzword o' the day") and I don't have an answer for them. Full text indexing: something like google does for HTML pages, or fts.postgresql.org does for the PostgresSQL mailing list archives. Good full text indexing programs understand something about the language they're indexing and automatically deal with plurals, suffixes, etc. Full text indexing is useful for many applications that include text data along with other more structured information. It is painful to have to use SQL plus some other facility for searching; just for starters it usually means keeping a second copy of the data. I suspect the push for full text indexing in PostgreSQL is stronger now that the 8KB row size limit that "ancient" releases had is gone. > 2. I propose that future versions of PostgreSQL require a primary key at > table creation. Frankly, I'm a little mystified as to why this was not > done already, but it's not too late to correct ... SQL uses a "bag" model, not a "set" model. Duplicates are allowed. I doubt that PostgreSQL wants to step away from both the standards and current practice in this area. Regards, Giles