Peter Eisentraut <peter_e@gmx.net> writes:
> Tom Lane writes:
> As an alternative syntax I can suggest
>>
> SET name TO value [ ON COMMIT RESET ];
>>
>> Ugh. Why can't we stick with SET LOCAL?
> SET LOCAL is already used for something else in the SQL standard. Not
> sure if we'll ever implement that, but it's something to be concerned
> about.
Actually, it looks to me like the spec's SET LOCAL has a compatible
interpretation: it only affects the current transaction.
My main gripe with "ON COMMIT RESET" is that it's a misleading
description of what will happen --- RESETting a variable is quite
different from allowing it to revert to the pre-transaction state.
regards, tom lane