Re: Significantly larger toast tables on 8.4?

On Monday 05 January 2009 18:45:49 Alvaro Herrera wrote:
> I did some measurements months ago, and it was very clear that libz
> compression was a lot tighter than the PGLZ code.

Back to the issue at hand.  The question at the top of the thread was which of 
the following behaviors we'd like by default:

(1) Compress everything within reason by default, causing slower retrieval, do 
not offer substr optimization. [<= 8.3]

(2) Compress only up to 1 MB, causing faster retrieval, supporting substr 
optimization. [8.4devel]

I am personally completely puzzled by option number 2.  Is there even a single 
use case for that?