Somebody in the thread at some point said: >>> There are hundreds, perhaps thousands, of different WiFi devices. >>> It seems to me that the onus is on the developer >>> to ensure as far as possible that changes he/she makes do not impinge >>> on devices working with the previous version. >> Well that's not going to happen, because the only way someone writing >> code can meet that standard is to sit on his hands entirely. > > Please read what I said, namely "as far as possible". *shrug* With something like wireless no one developer has the set of "devices working with the previous version". Either they sit in a funk quivering and asking themselves "Is it conceivable that this change may not work with some device somewhere in the universe?" before every edit as you recommend or they get on with improving matters as best they can. > First of all, drivers for new devices can be offered as new modules, > in which case they won't impinge on anyone not opting to use them. > > Secondly, I would imagine that judicious use if #ifdef's would ensure > that new code would not impinge on working devices. Over time the 'judicious use of #ifdefs' turns code into a nightmare temporal paradox as holistic as a box of broken safety glass. >> It also presupposes that the existing version of the code is in a golden >> state that any deviation from will make matters worse. If the existing >> code is shaky and shows no signs of maturing by small tinkering, then it >> has to be ripped out by the roots and redone. That totally violates >> your concept above, yet is the right thing to do. > > In the very rare cases where this is done, > the old module is normally left as an option, > and the replacement is described as EXPERIMENTAL, at least for a while. Not at all, there is constant churn in the kernel in areas that have much smaller granularity than "a module". APIs are broken and wholesale reimplemented all the time. >>> I must say, it's not really clear to me >>> why something like WiFi should actually change >>> between Fedora kernels corresponding to the same vanilla kernel. >>> Is the Fedora community regarded as a test-bed for such development? >> Several wireless drivers are in Fedora kernels that are not in vanilla. >> What do you prefer, no support for those devices at all or the best >> support that is currently available, so people can feed back issues and >> make the support better? > I'm all in favour of new drivers (as modules) > being as widely available as possible. > That should not impinge on anyone who does not opt to use them. Well then, turns out you are in favour of the question you originally asked, I guess that's the end of the thread. -Andy
