On 13 Jun 2009, Lennart Poettering outgrape: > On Sat, 13.06.09 17:09, Nix (nix at esperi.org.uk) wrote: >> In that case the likelihood of PA ever becoming a serious >> infrastructural component are nil. > > Aha. Is that so. > >> Imagine the X or libc developers said the same thing! > > Come on, this is bullshit. Complete and utter bullshit. > > Remember libc4, libc5, glibc? Yeah. One breakage in the last twelve years, and a statement that there will be further breakages over the maintainers' dead bodies. > Remember xlib vs. xcb? xlib still works. xlib wasn't broken by the introduction of xcb. > Remember the gtk1, gtk11, gtk12, gtk2 jumps? The reason why this was done in huge jumps was because the API of gtk2 has *remained back-compatible*. Rare, giant breakages are annoying but tolerable (the old system still works). What's not is a constant trickle of tiny breakage: then it's impossible to come up with something that works reliably against them. See ffmpeg for a terrible example. > We are going through transitions all the time, all across the > landscape. And yes, eventually there will be one in PA land too. So > what? Every time they happen they annoy all your users. If they happen too often you find you don't have any left. ABIs are important.
