The recurring plan that seems to be developing is to maintain the current driver focusing on it's stability and compatibility. Going forward the energy and focus would be towards a new driver which is liberated from the concerns of backwards compatibility.
Uh, what? Which backwards compatibility were you intending to throw away? Just older JRE versions?
Long term we'll have to see whether we want to maintain two codebases. For now let's experiment and see how things progress. There are pros and cons to both codebases, neither are perfect. With that in mind this is open source and people are free to do their own thing but there is strength in numbers. We obviously all care about PostgreSQL and it's JDBC driver so let us all try and make the best JDBC driver(s) possible!
As a long-time user of the driver, I'd like to de-lurk to point out that the nervousness about stability which has surfaced in this thread (and which I share) can often point to opportunities to improve automated test coverage.
Dave, how complete would you say that the existing test suites are? If a new implementation can pass the tests, how much does that say about the driver's completeness or backwards compatibility?
And for those who have forked the driver or started from scratch, do you have comments on the existing test suite? Or have you written your own?
If the AIO etc driver is to be developed in parallel with the existing one, it might be an interesting idea to split the tests out into a separate project so that the tests can be run against both drivers. Then improvements in the test suite are shared as well!