[Toybox] complaining about 'ifconfig' - better use 'ip'

[Rob Landley]

On 04/04/2013 09:52:47 AM, Isaac Dunham wrote:
> On Thu, 4 Apr 2013 10:17:17 +0200
> Bastian Bittorf <bittorf at bluebottle.com> wrote:
> > > >if you are
> > > >really using 'ifconfig', you should better switch now. the
> > > >pain will be much worse if you wait again 2 years...
> > >
> > > In what way? I vaguely recall first hearing about "ip" back under  
> Red
> > > Hat 9, which shipped in 2003. So I've been successfully largely
> > > ignoring it for 10 years now. What's queued up to change in the  
> near
> > > future?
> >
> > i will. there is no work done anymore on ifconfig, even the manpage
> > says to switch.

Noting, in passing, that he once again didn't answer my actual  
question. Unless "i will" is an answer to "what's queued up to change  
in the near future?" But this entire thread has been like that...

> Well, if toybox implements ifconfig, there will be work done on  
> ifconfig.
> I won't be switching.  And Rob won't be.  And Android won't be.
> 
> In other words, if you're switching, who cares?  If you want an ip  
> command,
> please either write it yourself or pay for it to happen. It's not on  
> Rob's
> agenda at present, and this argument over whether to implement it or  
> ifconfig
> is making no headway and impeding real progress.

This is what I popped in to answer, in a thread I've otherwise largely  
lost interest in.

It was never about ifconfig _or_ ip. It was about implementing an ip  
interface in addition to the ifconfig that somebody already submitted  
and that toolbox already has.

Support for ifconfig was in the first version of the roadmap I checked  
in a year ago:

   http://landley.net/hg/toybox/rev/639

Support for "ip" still isn't. And nobody other than Bastian's ever  
asked for it. (They've asked for things like "freeramdisk" and  
"readahead", but not that.)

> > every developer in the network world has realized, that doing the
> > 'ifconfg/route/arp/netstat'-thingy was a dead-end-street.
> 
> As far as I can tell, that problem is only with the net-tools  
> implementation of
> ifconfig. BSD still uses ifconfig, and so does Android.

What exactly is the _problem_? The existing ifconfig works fine. This  
is like saying http://nc110.sourceforge.net/ is useless because 1.10  
came out in 1996, and it's SO useless that I shouldn't have written a  
new implementation of for busybox. Which Denys then _replaced_ with the  
1.10 version from 1996 as the default, back before I relaunched  
toybox...

   http://lists.uclibc.org/pipermail/busybox/2010-June/072734.html

And I can see why he did it. (I don't agree, but I see why. I would  
have extended my version to have the features and optionally the user  
interface the other one had. But I wasn't really around for him to  
delegate that to, was I?)

But this "Oh no, the code is old! Code RUSTS!" I don't get it. Linux  
2.4 doesn't support modern hardware, containers, tickless operation,  
sucks at threading and SMP. But if it had decent arm support I expect  
android would have been based on it because it ran in a quarter the  
memory the current one needs.

Rob