IP over InfiniBand (IPoIB) Architecture
draft-ietf-ipoib-architecture-04

[Ballot comment]
Margaret and Vivek have shown a draft of the 04 revision which makes clear
that the IPv6 address use is a format.  It no longer mentions "valid IPv6 addresses".
It doesn't quite describe the allocation process to the extent of stating that GIDs
are not allocated by RIRs from the global IPv6 space, but there is no reason to
imagine that they are from this document.  I've removed my Discuss based on
the pre-submission 04.

[Ballot discuss]
Explain that the addresses are not "valid IPv6 addresses" in that the
infiniband forum does not intend IB sites to ask an IPv6 address registry
to allocate them from global IPv6 address space.  This is
undoubtably fixable with an RFC Editor note by changing/removing the
word valid and stating in the places where the term "valid IPv6
address" appeared that there was no intention to use addresses
from the global IPv6 address space including currently unallocated
IPv6 address space.  My point here is that even though huge, IPv6
address space, unlike EUI (IEEE MAC addresses) is not intended
for media-level addressing.  I am eager to resolve this Discuss very
quickly.

[Ballot comment]
Remove the document references from the Abstract.

  Remove the discussion about site-local addresses.

  Should there be a question mark in the title of section 3.3?

[Ballot comment]
It's nice to see a link layer that's designed for IPv6!

The paragraph mentioning site-local addresses should probably be modified to note that they've been deprecated for IPv6.