no easy answer (was Re: [PLUG] MySQL v. PostgreSQL -- perhaps an easy answer)

[Felix Lee]

Kyle Hayes <kyle at silverbeach.net>:
> That's very far from how I read this.  I asked about it on the Postgres
> list... hmm, two years ago?  Maybe less.  The answer I got was that Postgres
> did not have a consistent disk image until it was down.  You could recover
> from what was on disk (otherwise recovering from crashes would be a bit
> difficult, no?) but you'd lose data doing the recovery potentially.  That
> potential makes this method an emergency backup method at best.

maybe this was before postgres implemented write-ahead logging?
it looks like WAL was added in 7.1, released 2001-04-13.

"before WAL, PostgreSQL was never able to guarantee consistency
in the case of a crash."
    http://www.postgresql.org/docs/current/static/wal.html

with WAL (and fsync on), I think it should always be possible to
recover all committed transactions from a snapshot (or crash).
if not, it's a bug.
--