On Wed, 28 Jul 2004, Matthias Scheler wrote:

> In article <20040728110359.A08B71AE98@berkshire.research.att.com>,
> 	"Steven M. Bellovin" <smb@research.att.com> writes:
> >>> i can't expect whole IPv6 network will have MTU=1280.
> >>My outer IPv6 interface uses a MTU of 1280, too, and I can download files
> >>via FTP without problems.
> > In fact, the designers of IPv6 assert that you can make that
> > assumption.  It's a requirement for IPv6 links.
>
> I think Wojciech Puchar's concern is that other IPv6 systems may use bigger
> MTUs which breaks his tunnel.

IPv6 have path MTU discovery, and nodes should figure out how large
packets that can be sent, and not send larger.
The minimum MTU that IPv6 nodes *must* support is 1280.
So, unless you want to send larger packets than 1280, you don't need to
handle the path MTU discovery stuff, since 1280 is guaranteed.
If you want to send larger packets, you must do path MTU discovery.
Intermediate routers will not fragment packets in IPv6, so you will fail
miserably if you just blindly sends larger packets.

All of course just from the top of my head, so I might remember things
wrong right now...

	Johnny

Johnny Billquist                  || "I'm on a bus
                                  ||  on a psychedelic trip
email: bqt@update.uu.se           ||  Reading murder books
pdp is alive!                     ||  tryin' to stay hip" - B. Idol