Re: [Ipoverib] IP over IB DHCP draft last call

FWIW, and speaking as a regular WG member now, and having gone
reviewed both draft-ietf-dhc-3315id-for-v4-04.txt and
draft-ietf-ipoib-dhcp-over-infiniband-09.txt, and having gone and
reviewed some previous the threads on this topic...

The current document says:

    According to [RFC2132] the "client-identifier" option may consist of
    any data that uniquely identifies the interface. Examples of
    suitable "client-identifier" values are the link-layer address,
    fully qualified domain name (FQDN) or, an EUI-64 value associated
    with the interface.

In other words, any valid client-identifier can be used.

I see no reason why the IB document needs to do anything different
than just follow the recommendation in the 3315id. If there is any
reason why some other variant should be allowed, I'd like to hear
it. Leaving it up to the implementor to decide just creates confusion
and unecessary variation that serves little purpose (at least as far
as I can tell). Morever, the motivations for using a 3315id-style
client-identifier are compelling. So, unless there is a good reason
why IB can't use 3315-id, I'd suggest replacing the above paragraph
with something like:


   When using DHCP over an Infiniband interface, a client-identifier
   as described in [draft-ietf-dhc-3315id-for-v4-04.txt] MUST be used.

Thomas