Packetization Layer Path MTU Discovery BOF (plpmtud)
Thursday, Match 20 at 1530-1730
================================
CHAIR: Matt Mathis <
[email protected]>
DESCRIPTION:
This BOF is to introduce a new "Packetization Layer Path MTU Discovery
Algorithm" and discuss IETF venues that might be used to complete the work.
This new algorithm does not rely on ICMP or other messages from the network (so
it is not subject to the problems described in RFC2923). Instead it finds
the proper MTU by starting with relatively small packets and searching
upwards by probing with test packets.
Key portions of the algorithm need to be implemented in the transport layer.
In particular any lost probes should not be treated as congestion signals. For
this reason, it can be argued that this work belongs in the TSV area.
Many of the unresolved design issues pertain to the efficient use of shared
state between concurrent and successive connections to the same destination.
The complexities of the this aspect of the algorithm are due to the huge
technological breadth and subtleties of the delivery semantics associated with
the network and lower layers. For this reason, it can be argued that this work
belongs in the INT area.
The ID draft-mathis-plpmtud-00.txt describes a preliminary version of the
algorithm. You can join the list be sending a note to
[email protected] with
"subscribe mtu" or "subscribe mtu <alternate mailbox>" in the body. The
list itself is
[email protected].
The webpage at
http://www.psc.edu/~mathis/MTU/ describes two independent
prototype implementations and other related work.
Agenda:
Background (40 Minutes total)
- Quick overview of the problem
- Quick overview of the work-to-date
- Inventory technologies harmed by RFC1191 problems.
- Quick introduction to open issues
How can we best complete the work? (20 Minutes)
- Is this issue currently relevant to the IETF?
- Is this work tractable within IETF?
- How do we consider the operational impact?
- How do we consider deployment stratiegies?
IESG contacts:
Allison Mankin <
[email protected]>
Scott Bradner <
[email protected]>
Thomas Narten <
[email protected]>
Erik Nordmark <
[email protected]>
Key participants:
Matt Mathis <
[email protected]>,
John Heffner <
[email protected]>,
Kevin Lahey <
[email protected]>,
Peter O'Neil <
[email protected]>,
Fred Templin <
[email protected]>