"Dave Page" <dpage@vale-housing.co.uk> writes:
> I think the first thing to consider, is whether it would be easier to
> accommodate point 4 as well (upgrade to v3 protocol) at the same time as
> adding SSL support by ripping out all the comms code and replacing it
> with libpq. I'd really like to hear some opinions on that idea before
> anything else is done, as not only would it solve both the current
> problems, but it would also future-proof us against future protocol
> changes.
> The best bet for this one is simply to start looking at the code and try
> figuring it out I think - I warn you though, this one is probably not
> gonna be easy
One point to make here is that libpq doesn't currently expose the full
power of the v3 protocol, and it's entirely possible that there are
things ODBC needs that you just can't get at in current libpq. We'd
certainly be open to well-designed extensions of libpq's API to fix
any such problems ... but keep in mind that if such is needed, you'd be
tying ODBC to 7.6 or later libpq. Might pose some distribution
problems.
I think that reimplementing ODBC on top of libpq would probably be
a good plan in the long haul, since in theory it ought to reduce
maintenance effort in the long haul. But it won't be a quick fix.
regards, tom lane