Re: comments on mech-v2-01

["Chirayu Patel" <chirayu@chirayu.org>]

On Tue, 11 Nov 2003 15:57:41 +0100 (CET), "Erik Nordmark"
<Erik.Nordmark@sun.com> said:

> > > The reason it was written like this was to not immediately
> > > invalidate any older implementations that might have sent the link-
> > > layer address options. I don't know if there ever were any such
> > > implementations.
> >
> > Do you think we should change it now? I prefer "SHOULD NOT" and
> > "SHOULD ignore".
>
> Problem is if both ends of a pair of nodes don't satisfy the
> recommendation stated in the SHOULDs. In that case things would not
> interoperate. Thus I think there has to be a MUST applied to at least
> one of the sender and the receiver.

Hmmm... You are right. So, it has to be either 1) "MUST" and "SHOULD
ignore" (as it is today), or 2) "MUST", and "MUST ignore". Probably we
should not care that much about invalidating older implementations (if
any) as this is an easy requirement to implement, and go with the later.

CP