Hacker News Reader

Maybe there are some core PostgreSQL hackers here:

I know this probably sounds silly but for the transaction ID thing, it does seem like a big deal, is it really insurmountable to make it a 64 bit value? It would probably push this problem up to a level where only very, very few companies would ever hit it and from a (huge) distance the change shouldn't be a huge problem.

View on HN

There have been several discussion about this and I if I recall correctly the main issue is that this would bloat the tuple size even more (PostgreSQL already has a large per-tuple overhead). The most promising proposal I have seen is to have 64-bit XIDs but only store the lower 32-bits per tuple but have a per-page epoch for the upper bits.

store it (something) like protobuf does - the smaller the number, the less bytes it takes?

Reply on news.ycombinator.com