On Wed, Mar 26, 2025 at 9:14 AM Nathan Bossart <nathandboss...@gmail.com> wrote:
> FWIW one of the big reasons I didn't proceed with the enum approach initially is because I worried that I'd end up in a similar discussion > about how terrible _that_ approach is. When I look at that patch [0], I > genuinely wonder if folks would accept that without the isset_offset > context. Maybe I misjudged... > > I think this discussion was going to happen no matter which approach was actually committed. The concept of "is set" is too obvious and clean a solution to not be brought up and considered; while at the same time this argument about staying consistent with nearby code, even in the face of a hack-ish implementation, was going to need to be explicitly considered. This discussion was preordained the moment we decided to add vacuum_truncate to the system. It was only a matter of when. And while hindsight is 20-20 your own comments regarding your uncertainty suggests you at least had an inkling of suspicion that this discussion was going to be part of the outcome of committing this. I would have been fine with: "I committed this approach because it's cleaner, and here is why I dislike the enum approach. Let's have a discussion in May if this choice is unappealing for reasons." Getting in the user-facing feature "DBA choice of default" before feature freeze was warranted and the patch as committed did meet all the necessary requirements. David J.
