> While I agree with 99% of what Roy said, I would like to make one > clarification from my point of view. I don't really care if it conflicts > with a name on Unix. I do however care if the name conflicts with > something in POSIX or something in common use on multiple platforms. If > we implement something that conflicts with a name on Windows, and our > implementation doesn't look incredibly similar to that entity on Windows, > then I consider that a problem.
ack. it _is_ the NT api i'm after. the whole lot, but at this time, just the NTNamedPipes ones. really, i _really_ think tht that apr nt version should just be v.clearly a shim directly to TransactNamedPipe and CreateNamedPipe, and then the horrors - that _don't_ involve unix namedpipe AT ALL but unix domain sockets instead - go on to emulate that under unix, with APR_NOTIMPL returned for bits decided _not_ to be implemented right now. and specifically, right now, that means, in my book, if the name contains a remote server on unix, you return APR_NOTIMPL. what you think? lukes
