On 08/07/2020 06:42, Stephen Hemminger wrote: > On Tue, 7 Jul 2020 02:03:36 +0300 > Pavel Begunkov <asml.sile...@gmail.com> wrote: > >> On 07/07/2020 01:28, Steven Rostedt wrote: >>> On Tue, 7 Jul 2020 01:17:47 +0300 >>> Pavel Begunkov <asml.sile...@gmail.com> wrote: >>> >>>> Totally agree with you! But do we care then whether two _devices_ or >>>> _objects_ >>>> are slave-master? Can't see how it fundamentally differs. >>> >>> The term slave carries a lot more meaning than subordinate. I replied to >>> someone else but later realized that the person sent me their reply >>> offlist, so my reply to them was also offlist. What I told them was, >>> back in college (decades ago), when I first mentioned "master/slave" in >>> conversation (I think it was about hard drives), a person in that >>> conversation stated that those were not very nice terms to use. I blew >>> it off back then, but after listening to more people, I found that >>> using "slave" even to describe a device is not something that people >>> care to hear about. >> >> That's cultural, but honestly I've never seen such a person. I still >> don't understand, why having secondary or subordinate object belittling >> the owned side by not providing it the same rights and freedom is OK, >> but slave/master objects are not. Where is the line? >> >> >>> >>> And in actuality, does one device actually enslave another device? I >>> think that terminology is misleading to begin with. >> >> As mentioned, I do like good clear terminology, and if it conveys the idea >> better, etc., then it's worth to try. And IMHO that's the right reasoning >> that should be behind. Otherwise, for almost every word we can find a person >> seeing something subjectively offensive or at least bad in it. > > Wherever possible the kernel should use the same terminology as the current > standard in that area. Many of the master/slave references in the networking > code are for protocols based on IEEE 802 standards (unfortunately paywalled). > The current version of those standards do not use this kind of wording and the > standards committees are also actively working on inclusive language > statemets. > > As far as the use of master/slave for bonding, bridge, team etc, it > looks like Linux just invented using those terms since I don't see it > any other vendors implementations Cisco/Juniper/Arista/... Linux terms > are different than industry norms in networking, this is not a good > thing. But changing human expectations is hard.
And that's a perfectly convincing argument for a change -- consistency makes it easier to work with specs and code. I've never said anything against. I care about arguments being logically sound, as yours are. And the author neither provides such, nor IMHO actually helps the issues it raised. -- Pavel Begunkov
