Bruce Momjian <pgman@candle.pha.pa.us> writes:
> Tom Lane wrote:
>> [shrug...] I'd counsel leaving this as-is.
> What information are we loosing by having START and BEGIN use the same
> nodes? Knowing what keyword they used to start the transaction?
Exactly.
> Seems that would only be important if we wanted them to behave
> differently, which we don't, I think.
Whether we want them to behave differently or not, we need to preserve
the difference. The prior cases where the parser smashed two different
inputs into the same parse tree have all been "because it doesn't
matter", and sure enough we've usually eventually decided it did matter.
regards, tom lane