On Tue, 2007-01-16 at 14:53 -0800, Chris Darroch wrote: > Well, I'm not sure there's a policy, as such. At a minimum, I'd > think 10g and 9i should be supported. 8i (8.1, IIRC) would be nice too. > I seem to recall that 8.0 had somewhat incomplete features, since a > lot changed from 7; OCI got a big makeover. I'd be inclined to suggest > not trying to support anything earlier than 8i, myself.
OK. I'll keep that in mind. If you have any real life experience with OCILobRead() and how the character v. bytes counting *actually* works, it would be good to know. The manual is clear as mud on the issue and when building a bucket brigade, one needs to do repeated reads from a position where the previous read left off. The new function OCILobRead2() has much better semantics for this, but unfortunately only comes with 10g. The other solution is, of course, to just make one big bucket if OCILobRead2() isn't there. Brute force and all... > Another to-do item is adding version detection to the autoconf > stuff. As with so many things in Oracle, this is, I suspect, > non-trivial. DBD::Oracle's Makefile.PL function get_client_version() > is pretty terrifying. The stuff in APR-util's build/dbd.m4 is tame > by comparison. Yeah, I'm not sure I want to get into this one :-). BTW, I added detection of libnnz10 recently to dbd.m4, which should take care of linking of some flavours of 10g, I guess. <off-topic> Honestly, I find the whole OCI terrifying. IMHO, all open source databases we currently support have more elegant APIs. Not to mention better documented. <sigh/> </off-topic> -- Bojan
