RE: [pmtud] RE: [dccp] PMTU issues

["Phelan, Tom" <tphelan@sonusnet.com>]

Hi Itojun,

>> This conversation about UDP apps has inspireed me to specificlly cover
>> that 
>> case in the next iteration.  Has anybody tried to implement pmtud
>> using UDP?
>> I have a hunch that there are some API gotchas for current OS's.
>
>	UDP is a unreliable data transport, so loss of packet due to
> PMTU is
>	okay.  *BSD can do PMTUD for UDP without much change (set DF=1
> in
>	udp_output)

This is at the core of whether this issue is a problem or not.  The view
from the transport level is that this is an unreliable transport, so packet
loss is OK.  I think that the view of an application using an unreliable
transport is that if a packet is lost the app can live with it and it isn't
worth retransmitting the packet, but that it's much better if packets are
not lost in the first place.  With the current DCCP PMTU mechanim, if an app
wants to optimize the MTU it uses it's likely to lose some packets.  I think
that most apps wouldn't think that's a good tradeoff and would look for some
method that didn't routinely lead to packet loss, even if it were less
optimal.

Tom P.

_______________________________________________
dccp IETF mailing list: dccp@ietf.org
list info:  https://www1.ietf.org/mailman/listinfo/dccp
wg charter: http://www.ietf.org/html.charters/dccp-charter.html