Subject: Re: CVS commit: src/sys
To: Robert Elz <kre@munnari.OZ.AU>
From: Jonathan Stone <jonathan@dsg.stanford.edu>
List: tech-net
Date: 12/08/2004 11:36:30
In message <26418.1102480278@munnari.OZ.AU>Robert Elz writes

[...]
>No comment.    But assuming that is correct, so what?   Lots of code has
>bugs, the answer to that is generally to fix the bugs, not to discard the
>code.

I'm highly skeptical that multiple-loopback can be made to work right
*at all*.  Especially, since I looked at the loopback mtu-change code,
how it interacts with the routing code, and all the places in Peter's
patch which are using lo0 (now via lo0ifp), with no regard at all to
specific addresses.  So I didn't look 11 years ago; so sue me.

If you want to say we should be having that discussion independently
of making if_loop a cloner, I don't agree:


>If there are bugs in loN for N > 1, then let's find and fix them.
>If.

Not ``if''.  The semantics don't match the various uses that've been
suggested here as justifications for cloning loopback. That's a fact.

>
>  | Diving in and making the config-time->dynamic
>  | change (with an explicit rejection of any discussion) makes it very hard
>  | to achieve that.
>
>No, in this case it was the right way.

Robert, that's just nonsense.  Remember ``no code before its time''?
(I'm sure Christos does.)  There was absolutely nothing preveting
Peter Postma from fixing any issues with clonable loopback interfaces
*BEFORE* committing the cloning changes. That's clearly a better way
to proceed.