RE: [pmtud] I-D ACTION:draft-savola-mtufrag-network-tunneling-00.txt(fwd)

[Pekka Savola <pekkas at netcore.fi>]

I've tried to make the host-to-router case out of scope for this spec, 
but this seems interesting, so pursuing..

On Fri, 18 Jun 2004 Mika.Joutsenvirta at nokia.com wrote:
> In host-to-router case, where application wants to 
> send > MTU UDP packet to server behind router. 
> In common case IP stack generates UDP packet and then
> starts to perform tunneling. If host now follows rules
> from 3.1, it first encapsulates (IP-in-IP, IPSec, etc.)
> this big UDP packet and then fragments it to suitable
> pieces. [...]

To be clear, this happens when the MTU value for the IP-in-IP or 
similar tunnel is too high, correct? 

For example, assume that the MTU for IP-in-IP tunnel is 1480 bytes
(reasonable in many cases).  UDP packets bigger than 1480 bytes would
be fragmented or denied before encapsulation.  On the other hand,
let's assume you'd send such packets of 1480 bytes:

 1) the packets are encapsulated fine (total 1500 bytes)
 2) if the router's egress link is smaller than 1480 bytes, it will
send (if the inner packet was v6 or v4+DF set) ICMP packet too big
message back to the source.
 3) if the source can do PMTUD (it should!), it decreases the PMTU
value for destination, down from 1480, and this problem is avoided for
subsequent packets.

(On the other hand, if 1500 was used instead of 1480 -- typical -- 
you'd also have:
 1.1) the outer packet would probably get fragmented
 1.2) the fragments need to be reassembled by the router
+ adjust the other values accordingly)

Is there something I'm missing here?

Or are you assuming that the tunnels would have "infinite MTU" (or
64K) -- i.e., that the inner packet would never get fragmented, as
described (as a forbidden approach) e.g., in section 3.2 (couple of
first paragraphs) of draft-ietf-v6ops-mech-v2-03.txt ?

-- 
Pekka Savola                 "You each name yourselves king, yet the
Netcore Oy                    kingdom bleeds."
Systems. Networks. Security. -- George R.R. Martin: A Clash of Kings


_______________________________________________
pmtud mailing list
pmtud at ietf.org
https://www1.ietf.org/mailman/listinfo/pmtud