First of all: this message is going out to the IPv6 wg and internet
area mailinglists (for reasons explained below) but it's not
crossposted, so replies will go to the list where you found it by
default unless you add the other one yourself.
The internet-drafts administrator was kind enough to publish the
following draft:
http://www.ietf.org/internet-drafts/draft-van-beijnum-multi-mtu-00.txt
The idea here is to lift the limitation that all nodes connected to a
subnet must all use the same MTU by adding mechanisms to negotiate
the use of bigger packets between two nodes. Hopefully, this will
make it possible to start increasing the MTU on many links that make
up the internet. Currently, there is some form of ethernet in almost
all paths across the internet, and all forms of ethernet use an MTU
of 1500 bytes, even though gigabit ethernet hardware generally
supports larger sizes. The problem is that switches can't fragment
and ethernet is stateless so the IEEE can't mandate larger packet
sizes for faster ethernet. But IP can fragment, or, more generally,
adjust the packet size, and has at least some per-neighbor state so
we can solve this at the IP layer.
Ultimately, the idea is that this will work for both IPv4 and IPv6.
So at a high level, this would be an appropriate discussion topic for
the internet area list.
However, IPv4 doesn't have many hooks where new options can be added
easily like IPv6, so I'm focussing on IPv6 at this time. The draft
specifies:
- a new router advertisement option
- two new ICMPv6 messages
- a new neighbor discovery option
So it would be nice to get feedback on the protocol details from the
people who tend to hang out in the IPv6 wg.
Any feedback is greatly appreciated.
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------