On 14/09/17 09:38, Dave Mill wrote:
And then send those packets through a TCP MSS adjuster. This works well for the rare case where the customer can't do it.
Really, the 1492 packet size is just a non issue.
I guess the Internet really is only TCP these days... and presumably soon only TCP/443 (maybe with a bit of legacy TCP/80). The other 120-ish IPv4 protocols are unused.... or should that be unusable. TCP MSS clamping is a clever hack to work around PMTU Discovery issues caused by MTU mismatches along the path. But it only works for TCP, and then only for unencapsulated TCP. It wouldn't be necessary if deliberate MTU mismatches weren't being designed into even "new" networks; 1500-byte MTU might only be a de facto standard (due to the ubiquity of Ethernet now), but it is a wide spread de facto standard. FWIW I'm personally more annoyed by PPPoE being unable to go through the fast path forwarding on CPE devices, and thus being CPU performance limited, than about the MTU reduction. But the lower user-visible MTU seems almost entirely avoidable -- and yet not avoided in the common case. Ewen PS: On the plus side, everyone's PMTU Discovery algorithms are getting a good work out.... :-)