At 11:16 AM -0700 5/24/00, Richard Draves wrote:
>The mobility draft has at various points or times talked about swapping the
>source address in the IPv6 header and the address in the home address
>option. This is OK conceptually and I understand that some implementations
>actually work that way. But at least in our implementation, we treat packet
>data as read-only. I would not like to see specifications that assumed the
>swapping really does happen.
There is precedent for address swapping while a packet is in transit:
the Routing Header works that way. Granted, no swapping happens at
the source or final destination node, in that case.
>Considering the case of a mobile node initiating a TCP connection to a
>global address, with a choice of a global home address and a global care-of
>address, I think the desirable default behavior to use the global home
>address for the TCP endpoint and insert the home address option (and a
>binding-update option in the SYN). I think this is the desirable default
>even if the care-of address and the destination address appear to be
>topologically closer than the home address and the destination address.
I agree. That way the TCP won't break if and when the mobile moves,
which is one of the main reasons for doing mobile IP in the first place.
It would also be good if common apps made it very easy for the user to
override the default behavior and force use of the COA address as the
source, for those cases when the user (a) knows that the destination is
"close" and (b) doesn't plan to move before the session ends or doesn't
care if a move breaks the session.
Steve
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page: http://playground.sun.com/ipng
FTP archive: ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------
