Scott Marlowe wrote:
> So, they did exactly what pgsql crew did with TOAST, but instead of
> sticking it under an existing datatype that everyone already had, they
> made another new type to keep track of. I can't think of a reason to
> not just update the text type to be just like the leader's,
> Postgresql's, but I'm sure they have their reasons. I'd love to hear
> them though.
Their text type was accessed through BLOB interfaces at the
application/SQL level, so just substituting it would break a lot of things.
That said, you'd think they could provide wrappers to allow code used to
using the BLOB interfaces to operate on the new no-longer-a-blob type.
Perhaps there are performance issues there (say, it being more expensive
to repeatedly re-write a tuple with a VARCHAR(MAX) field than to
re-write a TEXT field) that meant they preferred to separate it out.
Whatever the SQL server folks are, they're not stupid, and I can't
imagine they'd do this without at least a half-decent reason.
--
Craig Ringer