On Mon, Jan 26, 2004 at 11:04:47PM +0100, Jean Delvare wrote: > Hi all, > > The more I work with new sysfs names, the more I wonder why the naming > scheme was chosen the way it is. We had features named > <type><number>_<subitem> for years, and now end up with files named > <type>_<subitem><number>. Not only I don't see why a change was needed, > but I definitely prefer the old naming scheme. It was much more logical, > since "ls" would order them in a very logical way (grouping by type and > sub-grouping by "physical entity"). So can you give me an example of what you want to change? > What's more, the libsensors tweaks would be much more simple if we had > stuck to the old scheme. Most of the required conversions are just > moving from old scheme to new scheme with no other change. We had other problems with the old scheme (multiple values per files being the biggest.) > Since the sysfs interface is obviously still not stabilized and we are > still on early 2.6 kernels, I think we should consider moving back to > the "old" names for the sake of consistency, beauty and logic. > > Greg, don't look at me that way please. I am serious! Heh, sorry for not getting back to you on this, been real busy lately... > I know this is quite an important change, but precisely, we'd better > change before porting the missing 30 chip drivers than after. I know how > much work it will be - especially since I'll obviously be the one to > cope with it - still I think we should do it. > > Comments, thunder and rocks welcome ;) > Thanks. > > NB: May I know the name of the brilliant chooser of the new naming > scheme? Not to blame him/her (oh well... ;)) but to know the reasons of > his/her choice. That would be me. See the archives for what I propsed, and why I did it. Of course I can't remember what I said back then, and no one seemed to disagree... thanks, greg k-h
