Re: snapshot too old issues, first around wraparound and then more.

On Wed, Jun 16, 2021 at 11:06 AM Andres Freund <andres(at)anarazel(dot)de> wrote:
> > 2) (a) Some hackers want the feature gone so they can implement changes
> > without making those changes cooperate with this feature. (b) Bugs in this
> > feature make such cooperation materially harder.
>
> I think the a) part is a large problem. Primarily because it's so
> unclear what one exactly has to do where (no docs/comments explaining
> that) and because there's no usable test framework.

Right. This is what I meant yesterday, when talking about design
issues. It's really not about the bugs so much. We probably could go
through them one by one until things stopped being visibly broken,
without going to a huge amount of trouble -- it's not that hard to
paper over these things without anybody noticing. This is clear just
when you look at how long it took anybody to notice the problems we do
have. Whether or not that amounts to "just fixing the bugs" is perhaps
open to interpretation. Either way I would not be comfortable with
even claiming that "fixing the bugs" in this way actually makes the
situation better overall -- it might make it even worse. So in a more
fundamental sense it would actually be really hard to fix these bugs.
I would never have confidence in a fix like that.

I really don't see a way around it -- we have to declare technical
debt bankruptcy here. Whether or not that means reverting the feature
or rewriting it from scratch remains to be seen. That's another
question entirely, and has everything to do with somebody's
willingness to adopt the project and little to do with how any
individual feels about it -- just like with a new feature. It does us
no good to put off the question indefinitely.

--
Peter Geoghegan