>> 	PMTUD is designed with assumption that link MTUs are stable (do not
>> 	change over time).
>So the simple approach (limiting to something a bit higher and leaving 
>figuring the optimal value to the admin) is the way to go?
>Would 1500-sizeof(struct mobile_h) be a reasonable upper limit for the MTU or
>should we make that basically unlimited (i.e. 0xffff), relying on the admin
>to know what he is doing?

	since GRE header is variable length, hard to tell.  i would set
	the default to whatever value commonly used by vendor routers
	as default (like cisco).

	as for upper limit, I see no reason to enforce upper limit as outer
	IP header has DF=0 (big frames can go through, though performance
	may suck due to fragmentation).  again, check what vendor routers do.

itojun