Alan McKinnon <alan.mckin...@gmail.com> writes: > On 26/12/2016 20:35, lee wrote: >> Tom H <tomh0...@gmail.com> writes: >> >>> On Fri, Dec 23, 2016 at 9:07 PM, lee <l...@yagibdah.de> wrote: >>>> Tom H <tomh0...@gmail.com> writes: >>>>> On Mon, Dec 19, 2016 at 3:07 PM, Daniel Frey <djqf...@gmail.com> wrote: >>>>>> >>>>>> It is even more frustrating that these so-called predictable network >>>>>> names actually can change on a reboot, it's happened to me more than >>>>>> once when multiple network cards are detected in a different order. >>>>> >>>>> >From Kay Sievers in [1]: >>>>> >>>>> <BEGIN> >>>>> Btw, predictable means it will not change between reboots, that names >>>>> will not depend on enumeration order within the same setup. It does >>>>> not mean or promise, that added kernel/driver/firmware features will >>>>> not result in different names. That is expected behavior. >>>>> </END> >>>>> >>>>> [1] >>>>> https://lists.freedesktop.org/archives/systemd-devel/2015-October/034614.html >>>> >>>> So the names will not change when rebooting and are to be expected to >>>> possibly change at any time. >>>> >>>> How is that more reliable? >>> >>> It's more reliable than using the kernel's names because the names >>> won't change UNLESS there's kernel/driver/firmware change for that >>> NIC. I doubt that these changes occur that often. Perhaps someone else >>> knows. >> >> What happens more often: That a network card is replaced with a >> different one or that the software changes? >> > > > OK, let me try explain this again. > > NIC names are tricky, several posters (myself included) have laid out > various methods and options by which it can be done. Experience shows > that in real life the simple traditional names are easy to remember but > prone to changing and (worse) prone to race conditions. Other methods > change less often in reality but the names are somewhat trickier to > remember. > > Opinions on these things differ; experience on these things differ and > people's use cases on these things differ greatly. A coder working in > this area has to decide what sort of cases they want to support, what > problems they want to attempt to solve and what new features they want > to introduce; then they have to write the code. > > Once the code is written, the coder then has to decide what nomenclature > to use when describing the software and the effects it has. In this case > centered around systemd a word was chosen: "reliable". > > Some will think it's a good name, some don't care, some will think it's > a bad name; and all of those things are basically irrelevant because the > name doesn't tell you much abut what the software will do. Reading the > fine manual will tell you that. It's all a part of being human because > our languages are imprecise, heavily overloaded and hugely redundant. So > are our spellings. But we are stuck with it because that's the general > emergent behaviour of a homo sapiens brain. > > Arguing abut this is about as nonsensical as arguing about whether "lee" > is a good handle on a forum or not. To a pedant it's a bad name, one > can't tell if you are male, female or if it's actually an Asian family > name.... > > Or one could do what most folk do, and not see a problem with 3 letters
I agree. What I don't agree with is that unrecognisable names generally make things easier (though they can, depending on the circumstances).
