Re: 2.1.120 - too many errors on the network interfaces

Brandon S. Allbery KF8NH (allbery@kf8nh.apk.net)
Sun, 20 Sep 1998 09:57:18 -0300

In message <19980920061601.20519.qmail@mail.ocs.com.au>, Keith Owens writes:
+-----
| On Sun, 20 Sep 1998 16:54:39 +1200,
| Chris Wedgwood <chris@cybernet.co.nz> wrote:
| >[about ifcconfig errors]
| >This is such a FAQ I'm beginning to think the we need to make sure
| >existing versions of net-tools (and similar things) check the output
| >is pretty much exactly what we expected - no extra fields, etc. and
| >barf if this is not the case.
|
| That's only going to be possible if *every* utility that uses ioctl or
| /proc can somehow decide that the interface format has changed. IMHO
+--->8

Dunno about ioctl (I've seen lots of schemes proposed and used, and all of
them fall short), but it does seem silly that proc doesn't include version
numbers. Would it really be that painful to have the first line of a /proc
file be an interface version number? Old utilities that don't know about it
would have to be replaced, but right now that ends up being true anyway and
with no warning other than a paragraph buried in some text file. Worse, if
you have to dual-boot between kernels with incompatible /proc file formats,
you can't easily deal with both from a single utility: but with version
numbers, a utility could invoke code conditionally on the version number or
even run a backend selected by version number (/usr/lib/backend.d/ifconfig/X.
Y, anyone?).

I'm not talking about anything really complicated here, just having
/proc/foo/bar start with e.g. "!@ 1.1\n".

-- 
brandon s. allbery	[os/2][linux][solaris][japh]	 allbery@kf8nh.apk.net
system administrator	     [WAY too many hats]	   allbery@ece.cmu.edu
electrical and computer engineering					 KF8NH
carnegie mellon university






-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/