2003-01-07T04:18:23 Darren Reed: > 2003-01-07T04:09:21 Balazs Scheidler: > Ok, so what you're saying is that you're wasting time by implementing > a library to an interface that has not yet been finalised. > > As nice and easy as it is to cut code, "don't do that". > > Get the detailed design right, THEN cut code. Whew. So never ever write code until you know precisely, for sure, what it should accomplish. Prototyping is bogus. I believe your position qualifies as "extreme". > Code isn't needed to critique a proposed API. If nobody has any experience in an area, a concrete implementation can be a big help for figuring out whether there are any unanticipated pitfalls behind a proposal. Sometimes you want to solve a problem even when you aren't 100% clear exactly what the ideal solution should be. A standard attack in the field of software engineering is to mock up a trial solution. If you get very, very lucky, perhaps it might end up fitting the need well enough to finish the job. More commonly, attempts to use the trial implementation reveal defects, which can be used to improve the specification of the desired solution. This starts a cycle; use/improve/repeat. At some point, perhaps early, perhaps late, it makes sense to begin a commitment to refrain from casually destroying backwards compat, to freeze interfaces and invite public use in production; then you call it 1.0. > > I don't thing the version number alone means anything > > 'unprofessional'. And by the way the library was not fully > > publicized, it was a release for the log-analysis mailing list > > for review. > and it is an open, public mailing list. where we're undertaking to work out concrete details of a good API. Balazs' implementation was offered for us to see whether it's along the right lines or not. At least that was my take on it. -Bennett
This archive was generated by hypermail 2b30 : Wed Jan 08 2003 - 08:49:13 PST