Re: [DNA-BOF] Re: [mobile-ip] Announce: BoF Proposal: DetectingNetworkAttachment

[Thomas Narten <narten@us.ibm.com>]

gabriel montenegro <gab@sun.com> writes:

> Greg Daley wrote:
> > As noted by Jari, there may be two parallel problem
> > spaces for these systems.  

> Yes, agreed, conceptually, this is true. The difference
> I see is that for the v6 side of things, the issue
> has attracted much discussion and attention at the
> IETF. In this sense it is  a problem asking for a BoF.
> The v4 aspect side of things would be a BoF asking for
> a problem.

I disagree at least somewhat here.

The issue has come up in IPv4. DHC (for IPv4) is known to not be clear
enough about what to do when one (re)attaches to a network. There is
also lots of experience here now that would at least be BCP
material. Some protocol clarifications might be appropriate for DHCv4
too (e.g., changing MAYs and unspecified stuff into stronger
recommendations).

In zeroconf (which is focused on IPv4), the IPv4 LL work has decided
that one should configure either a regular address (i.e., one learned
from DHC or manually configured) or an autoconfigured LL address. But
not both simultaneously. Thus, for robustness, they want to know if a
global address is available (quickly) or whether a LL address should
be configured. But the existing DHC spec isn't clear enough about
this...

So the issues are very much real in IPv4. I don't think we want to
ignore IPv4 and focus just on IPv6. But that also doesn't mean that we
can't have to parallel efforts given the differences in the underlying
technology.

Thomas