Re: Proposal for better support of time-varying timezone abbreviations

Michael Meskes <meskes(at)postgresql(dot)org> writes:
> On 15.10.2014 00:26, Tom Lane wrote:
>> * I've not touched ecpg except for cosmetic changes to keep the struct
>> definitions in sync, and to fix the previously-mentioned bogus free()
>> attempt. I doubt that it would be worth teaching ecpg how to access the
>> zic timezone database --- the problem of configuring where to find those
>> files seems like more trouble than it's worth given the lack of
>> complaints. I'm not sure what we should do about the obsolete timezone
>> abbreviations in its table.

> Maybe we should just remove thme for the new release. Yes, that might
> break some applications, but then the server doesn't know these either,
> so the applications might break anyway.

The same thought had occurred to me. Probably the main use of the
datetime parsing code in ecpg is for interpreting outputs from the
server, and (at least by default) the server doesn't use timezone
abbreviations when printing timestamps. So maybe that's largely
dead code anyhow. I would not propose back-patching such a change,
but we could try it in 9.5 and see if anyone complains.

A less drastic remedy would be to remove just those abbreviations
whose meaning has actually changed over time. Eventually that
might be all of them ... but in the meantime, we could at least
argue that we weren't breaking any case that worked well before.

regards, tom lane