Re: Commitfest problems

On Thu, Dec 11, 2014 at 1:26 PM, Josh Berkus <josh(at)agliodbs(dot)com> wrote:
> No, it's not a jump up by 2X, but it is an upwards trend. And I think
> that Tom has it right that the additional patches we're seeing are
> additional large, complex patches.

I feel like there's increasingly not that much easy stuff left to do.
Many of the problems we haven't solved yet are unsolved because they
are really hard, or because not that important, or both. For example,
if you look at the current CommitFest, there are things like:

Add ssl_protocols configuration option
alter user/role CURRENT_USER
Fix local_preload_libraries to work as expected.
Refactoring code for synchronous node detection
Refactor of functions related to quoting from builtins.h to utils/quote.h

Those things are all, obviously, important to somebody, or there
wouldn't be patches for them in need of review. But in the broader
scheme of things, they are very minor issues. And then you have
things like:

deparse DDL in event triggers
Compression of Full Page Writes
WIP: dynahash replacement for buffer table
Foreign table inheritance
Grouping Sets
INSERT ... ON CONFLICT {UPDATE | IGNORE}

Pretty much everything on that list is something we've been wrestling
with as a project for at least half a decade. I remember people
complaining about DDL triggers when I first got involved in the
PostgreSQL community, and the initial patch for foreign tables had
support for inheritance and constraints which I ripped out before
committing as too half-baked. Pavel had a GROUPING SETS patch by
2009. Brainstorming solutions to the full_page_writes problem has
been a perennial PGCon after-hours activity for as long as I've been
going. So it's natural that, to the extent these patches are making
progress, they're doing so slowly. It's hard stuff to get right.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company