On Sat, 9 Sep 2000, Zeljko Trogrlic wrote:
> First, having id and ID is a BAD naming practice.
True, but someone could have a database like that right now,
and we'd have to keep it working as separate identifiers for
some number of versions to allow people's dump files to be
restorable.
> Compatibility could be solved with compatibility flags. I guess this is not
> the only place where PostgreSQL breaks compatibility (like any other database).
I was more thinking about backwards compatibility.
I guess you could switch forward by having the database have two name
columns, one for the search name (lowercased) and one for the "real"
name. Then you'd need a hack that would lowercase the real name of
the identifier if you had a quoted and unquoted version with the
same case.
Cases such as ID, "id" would be failures (duplicate identifier) just like
it is now, Cases such as "ID", id would work (the quoted one also only
matches to "ID") and cases such as "ID", ID are the same as it is now.
