Mate Varga <m@matevarga.net> writes:
>> Using the large-object API for things that tend to not actually be very
>> large (which they aren't, if you've got hundreds of millions of 'em) is an
>> antipattern, I'm afraid.
> I know :( So maybe I need to do some refactoring in the application and
> inline the lobs. The data is binary data with very high entropy (encrypted
> stuff). Would you recommend bytea for that?
Yeah, it'd likely be less of a pain-in-the-neck than text. You would need
some sort of encoding anyway to deal with zero bytes and sequences that
aren't valid per your encoding, so you might as well go with bytea's
solution.
regards, tom lane