On 10/01/2015 07:55 AM, Tom Lane wrote:
> Playing devil's advocate ... would this really do much other than bloat
> the release notes? The entire assumption of this thread is that people
> don't, or don't want to, use the release notes to find out what got fixed;
> they'd rather search a tracker.
It's not a question of "rather", it's a question of how searchable the
release notes are, which is "not really at all". Yes, you can scan the
release notes for the latest update, but consider users who have an
issue and are running 9.2.7. Reasonably enough, they want to know that
their issue is fixed in 9.2.13 (or in 9.4 if it turns out to be a
feature, not a bug) before they ask their boss for a downtime. Figuring
that out now is really hard.
I tried to tackle this three or four years ago, by writing a tool which
would slurp the release notes and put them into a full-text search
database. This turned out to be very hard to do; our formatting for the
release notes makes it very difficult for an automated import program to
interpret (SGML doesn't help), especially on point releases to old
versions. It also turned out that the resulting database was useful
mostly to me, because you had to figure out what terms to search on
based on the bug report in front of you. As a result, it never went online.
So today, the only time the release notes are useful for a "is this
issue fixed or not" is when a release note message mentions the specific
error message the user is getting, which is a minority of the time.
So in addition to what Haas mentions, I think we want to be able to link
the release notes to the original issues for our hypothetical bug tracker.
--
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com
