On Mon, Nov 20, 2017 at 6:33 PM, Bruce Momjian <bruce@momjian.us> wrote:
>> That's not completely fair. Some of the items are actually implemented
>> without too much fanfare, but just never get removed from the Todo
>> list.
>
> I go through the TODO list after every major release and remove the
> completed items I find.
I was referencing the "sorting" section, which has some obsolete
items, like "Consider being smarter about memory and external files
used during sorts". To be fair, whether or not the other sorting items
are obsolete is more debatable, and it's evident that you've made an
effort to maintain it. It's just very hard for anyone to maintain a
todo list like this.
The idea that I was trying to express was my remark was that the todo
list inevitably becomes a thing with projects that are either likely
controversial, or inherently very difficult. Because if they weren't,
then they would have already been implemented (unless somebody forgot
to remove them!).
There is a lot of stuff like "Vacuum Gin indexes in physically order
rather than logical order". In theory, that should be totally
uncontroversial. In practice, it's very controversial, because the
reason that it doesn't already work that way is subtle, has a lot to
do with complicated concurrency issues. And, I suspect that the
existing GIN VACUUM code is buggy.
A list like the todo list sounds great, but in practice most of these
things are so high context that in practice the best thing to do is
seek guidance from experienced contributors. On the one hand, I think
that it's good for new contributors to choose their own projects, so
that they have a sense of ownership. On the other hand, there are
significant risks if they really do that. It takes careful
consideration on the part of the more experienced contributor giving
advice. I've only managed to give good project advice 3 or 4 times,
myself.
--
Peter Geoghegan
