At 04:06 p.m. 22/06/2005, Kevin Lahey wrote:

> > So... who wants to survey the modern Internet and publish a new RFC?
>
>For a solution to the problem of untrustworthy probes and just a general
>update of PMTUD stuff, see:
>
>         http://www.ietf.org/internet-drafts/draft-ietf-pmtud-method-04.txt

There are some problems with this approach. I raised those issues in the 
PMTUD WG mailing-list some months ago.

Some of them:
* You assume IPv4's minimum MTU is 576, and it is 68. This means that 
thatof "start sending small packets" is not an option.
* The draft assumes ICMP is evil. However, it can provide good information, 
and there are some checks and mechanisms you can implement to avoid them 
from opening the door to PMTUD attacks. I published a draft that describes 
all this (http://www.gont.com.ar/drafts/icmp-attacks-against-tcp.html).
* I'm not convinced about that of the "probing" if the assumed PMTU is 
correct.. If there's no multipath, the probe is not needed. If there is, 
the probes can succeed, and after that, packets could follow some other path.

Note: The mechanism proposed in my draft is not a replacement for a new 
PMTUD mechanism. It just avoids security problems with the current one. But 
some ideas in it (basically, the handling/validation of ICMP messages) 
should probably be incorporated in the PLPMTUD.

Kindest regards,

--
Fernando Gont
e-mail: fernando@gont.com.ar || fgont@acm.org