On Wed, Jun 12, 2002 at 05:48:46PM -0400, Bruce Momjian wrote:
>
> I can add it to CVS as interfaces/libpqxx and we can then let others
> merge your configure tests into our main configure. Let me know when
> you want it dumped into CVS.
Might as well do it right now, with 0.5.2. We'll call that 1.0, and
leave the more radical future plans for 2.0.
There are some things I'd like to do in future 1.x releases that will
affect the interface:- nonblocking operation, probably as a latency-hiding tuple stream;- change the way you select the
qualityof service for your transactor;- allow notice processors to have C++ linkage;- addtional bits & bobs like field
andcolumn iterators.
OTOH there's no point in delaying 1.0 forever I guess.
FWIW, I'm thinking of doing at least one of the following in 2.0:- an easy-to-use but intrusive object persistence
layer;- offload some of the work to BOOST if possible;- adapt the interface to be more database-portable.
But back to 1.0... Would it be a useful idea to also integrate my own
CVS history into the main tree? Or should I just keep developing in
my local tree and submit from there?
Jeroen