[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: dna bof report
Warning: A statement from an "implicated" person.
There are several issues that DNA is grappling with:
1. RFC 2131, ZEROCONF IPV4 LL draft conflicts and problems. There some
things that need to be fixed in both documents as well as conflicts between
them. For example, RFC 2131 as well as DHCPv6 says DHCP should be activated
on "Link Down" -- this should be "Link Up". Zeroconf IPV4LL document
suggests that a host that sleeps or senses "media down" should assume that
it is on an IPv4LL network and do claim/defend when it reconnects. If
implementors took this literally it would mean that a host using IPv4LL
addressing would not obtain a routable address again.
There is an immediate need to fix RFC 2131 as well as the ZEROCONF IPV4LL
document so as to address these problems. The proposal on the table was to
submit errata for RFC 2131, and fix DHCPv6 in author 48 hours, and to have
DHC WG review of the ZEROCONF IPV4LL document.
2. The basic model of DNAv4 and DNAv6 are quite similar, even though the
technologies differ. For example, in both case L2 "Hints" are processed;
there may be reachability detection phase; and there is an address
acquisition phase. Also in both cases there are issues with detecting
non-determinism; in IPv4 this arises when moving between private networks
(e.g. two gateways, each with address 192.168.0.1, on two different networks
-- will the host be fooled?), on IPv6 it occurs when testing reachability to
an IPv6 Linklocal address.
On this basis, the BOF participants leaned toward handling DNAv4 and DNAv6
in the same WG.
3. "The last WG syndrome". In IETF there is a tendency to view a BOF as a
once in a lifetime opportunity to get work done. So "consensus" tends
toward overload, on the theory that if Optimistic DAD isn't handled in the
DNA WG, then it will drop on the floor and take years to get done. In
reality, the IETF is more efficient than that, when we're doing our job.
My personal recommendation on DNA is as follows:
1. Try to figure out how we can address the most egregious issues in
DHCPv4/v6 and ZEROCONF right away, via the errata and doc change route.
2. Do a single document on L2 "hints" for DNA. Note that this is distinct
from the "trigger" concept in that DNA must be robust against misleading
"hints" -- where as triggers seem to imply more certainty (which of course
doesn't exist, which is a problem).
3. As part of the Charter review, ask the founders to come up with a Review
team to look at the DNAv4 and DNAv6 documents to make sure that the pitfalls
(and amount of work required) is well understood. These should be highly
knowledgeable reviews like Robert Elz, Keith Moore, and Brian Carpenter.
I'd suggest that this be done before a charter is written so that we avoid
"ZEROCONF syndrome" -- a WG formed without sufficient expertise in its core
area. If suitable "experts" cannot be found to generate the initial and
subseqnet reviews, then the conditions for the WG to succeed do not exist,
and a WG should not be chartered until the resources are available.
4. Handle Optimistic DAD in a separate venue. My recommendation would be to
create the equivalent of the TSVWG in the INT area -- INTWG, for the purpose
of handling small things that wouldn't otherwise deserve their own WG, like
Optimistic DAD. The idea is to create a WG where IP experts would
congregate so that we don't repeat the ZEROCONF fiasco, where we created a
WG focussed on making fundamental changes to IP that was populated mostly by
folks interested in building smart appliances. In other words, a mistmatch
between skills and the task at hand.
_________________________________________________________________
Help STOP SPAM with the new MSN 8 and get 2 months FREE*
http://join.msn.com/?page=features/junkmail