Re: [PATCH] Revert "cfg80211: make WEXT compatibility unselectable"

["Lennart Sorensen" <lsorense@csclub.uwaterloo.ca>]

On Thu, Jan 01, 2015 at 12:14:15PM -0800, Linus Torvalds wrote:
> So I'm not saying "ifconfig is wonderful". It's not.
> 
> But I *am* saying that "changing user interfaces and then expecting
> people to change is f*cking stupid".
> 
> The fact is, ifconfig is simple for the simple cases, but more
> importantly, a lot of people learnt how to use it. Saying "you should
> all change, because we made up a new syntax" is not good policy.

Perhaps it would be good to at least fix the buggy bits of ifconfig and
perhaps make it work with netlink instead of parsing /proc.  Maybe not
add features, but at least fix the broken bits and make it use modern
interfaces.  The interface to the user can stay.  And of course if it
happens to encounter a really old kernel, it should still remember how
to speak to it with the old interface.

> The people who did "ip" could have fairly easily have done a wrapper
> around the same code that also left the old "ifconfig" syntax. Then,
> distros could have trivially just dropped the old "ifconfig" package,
> and entirely replaced it with the new "ip" package.

Hmm, that might be a better idea.

> As it is, we have two different models, and they'll basically stay
> around forever.
> 
> For something like ifconfig, very few people care. But *all* the same
> arguments are true wrt "iw" and "iwconfig".
> 
> The people who are trying to deprecate the WEXT interfaces should put
> the blame firmly where it belongs - on the people who thought that
> "we'll just ignore all old history".
> 
> Because people who think that "we'll just redesign everything" are
> actually f*cking morons. Really.
> 
> There's a real reason the kernel has the "no regression" policy. And
> that reason is that I'm not a moron.
> 
> History matter. Legacy uses matter.

Sounds good to me.

-- 
Len Sorensen