In your previous mail you wrote:

   This is an interesting idea, and I think it is worth exploring.   

=> this is a typical IPv6 idea (:-)...

   However, there are a couple of issues to overcome...
   
   Would you perform DAD on these addresses before you use them?  Or  
   would you somehow delegate a prefix (perhaps longer than a /64) to  
   each xTR for this purpose?

=> IMHO the second solution (what I called a virtual link) is simpler.
Of course, it wastes resources so I agree with the "longer than a /64".
In fact, this depends on how many addresses are needed (and don't
forget one can play on both source and destination addresses).

   >  and not all BSD/Linux/MacOS/Windows boxes (software and hardware).
   
   I think that the idea is that the IPv6 UDP headers with zero checksums  
   will only be sent between the ITR and ETR, so it is only LISP nodes  
   that need to change.
   
=> you know it is not reasonable to introduce a third kind of
IPv6 nodes (hosts, routers and LISP-capable routers :-).

   However, I personally would like to be able to implement a compliant  
   LISP node (ITR and ETR) on a desktop operating system like Windows or  
   Unix, and I don't think that one can expect deployed desktop operating  
   systems to be updated (to allow zero UDP checksums in IPv6) any faster  
   than the estimates given for routing updates (3-5 years).
   
=> I share this concern.

   Also, the last time I dealt with zero UDP checksums (mid 1990s), it  
   was only possible to disable UDP checksums (for IPv4) in the unix  
   kernel, so they were on or off for the entire box.  Have things  
   evolved since then?

=> as far as I know the answer is no. The only different in modern code
is the handling of checksum offloading in NIC hardware.

   Is it possible to have IPv4 UDP checksums enabled  
   (both calculating them and checking them) for some sockets while they  
   are disabled (set to zero, and not checked on receipt) for other  
   sockets?  If so, what sockets option is commonly used for this in Unix  
   operating systems?
   
=> BSDs have only a system-wide sysctl, Linux offers a setockopt()
(SO_NO_CHECK). I don't know for Windows. IMHO, it is a bad idea to
disable UDP checksums, UDP-lite should be used instead.

Thanks

[email protected]

PS: about the original thread (0 checksums vs IPv4 to IPv6 translation),
DNS-OARC DITL should have the data.
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to