>>> Note that falling back to DNS updates means minutes of delay. > >> Yes, if they move simultaneously, they will get minutes of delay (due >> to DynDNS update), and the TCP session will break. > > That's not a problem. User's experience of the delay is, which is > why Mobile IP has servers of home agents as a stable point to > receive packets. > > Note that, as you assume NAT (not very elegantly at the application > layer), mobility must be port wise (which requires NAT transparency > at the transport layer), because access points may be behind NAT. > MOPS (Mobile TCP) - as defined in my spec, will work fine through source NATs (Home routers), assuming the Mobile Node is well... mobile. But it doesn't cover load-balancers (a.k.a. destinations NATs), and doesn't cover if mobile node client is behind NAT, but the server has moved (Home Router becomes a destination NAT, on traffic from Internet). -- -Alexey Eromenko "Technologov"