Re: About ext/sockets/

On Sun, 24 Jun 2001, Daniel Beulshausen wrote: >the issue that made the extension break was the step from resources to >longs for the socket fd's, this was necessary as win32 socket fd's are >different from bsd style socket fd's. So this is broken now only on *nix? Nice. Previously it worked only on unices, now it works only on win32? What an improvement. >as this change would already have broken the extension i decided to update >the api as well, mainly the functions names, but also some return types. Did you happen to think that maybe people have already got used to the way it works? And they have created lot of scripts depending on the old behaviour? Old behaviour was also consistent with how the low-level socket functions worked. Which makes it easier to those people who already know socket programming in C to use this extension. >the extension works pretty well under win32 now, but if there are functions >that are broken under posix i suggest that someone with more knowledge of >the bsd sockets than i has a look at them. Maybe you shouldn't have touched it at all then? You should have sent the patch first to the list. And after someone using *nix platform had tested and verified it really worked only after that you should have committed. Never heard this: 'test before commit' ??? >i'd vote not for changing the api back, rather than updating/extending the >yet available (tiny) sockets documentation. I vote we get the old api back. I don't care if it doesn't support win32. Document that instead. Or update the documentation then. And add working examples there. I'm not gonna waste my time trying to figure out this. --Jani