[Softwires] MAP and 4rd-U - how to try to converge

[Rémi Després <[e-mail]>]

Alain, Yong, Ralph,

 First, a little bit of history is IMHO useful to understand where we are:
 1.  At the Beijing meeting, end of September, I started, with Professor Xing Li, Wojciech Dec, and Congxiao Bao, a convergence effort based on requirements expressed by double-translation and encapsulation proponents respectively (ref www.ietf.org/mail-archive/web/softwires/current/msg02994). 
 2. In early October, Alain then entrusted Ole with the task to lead a team  having "to formulate a unified format to be used either in an encapsulation or double translation mode, for both hub & spokes and mesh models" (www.ietf.org/mail-archive/web/softwires/current/msg03024)
 3. Working on my own, looking for a stateless solution that would be more completely unified than just the address mapping, I found an approach for that, called 4rd-U. (Its reversible "header mapping" uses systematic IPv6 fragment headers, and TCP-checksum neutrality of mapped addresses. Ref draft-despres-softwire-4rd-u-00)
 4. During the Softwire meeting in Taipei, in his report as MAP-team leader, Ole had a slide stating that checksum neutrality would result into "destination spray" (i.e. would break TCP!). Time lacking for a technical challenge of such a claim, this prevented the WG from finding interest in pursuing the 4rd-U approach.  
 5. In the MAP-team meeting that immediately followed that of the WG, Ole obtained a rough consensus that features of the 4rd-U proposal (checksum neutrality and the V octet) would no longer be considered by the MAP team. 
 6. After that, Ole and I having discussed privately, Ole acknowledged on the WG list that his statement against checksum neutrality was technically invalid. 
 7. Some interest for looking more at the 4rd-U solution was then reported to the chairs, and Alain encouraged me to work on a revised 4rd-U draft for the next meeting (ref www.ietf.org/mail-archive/web/softwires/current/msg03296).
 8. On December 29, I then posted tools.ietf.org/html/draft-despres-softwire-4rd-u-02. Following comments I had received, and accepting that some needs weren't satisfied in draft-02, I posted on January 28 draft-03 which has two tunnel variants close to one another (4rd-E for encapsulation, 4rd-H for header mapping).
 9. On January 30, Ole announced to the WG that the 4 related MAP documents were available (ref www.ietf.org/mail-archive/web/softwires/current/msg03328).

 Thus, people involved in the two approaches have clearly done what they were asked to.

 Trying to converge between the two approaches is worth attempting because, as Ole recently wrote: "let us make it clear that these two solutions are solving exactly the same problem, and they solve it in the same fundamental way (A+P). the differences we're talking about here are what whistles, bells (and dongs) we want to add on to the base specification. consider it a buffet, any feature from one of them can be applied to the other."
 Now, the 4rd-U draft is IMHO the best starting point for a specification that suits implementors that didn't participate in the work, because it is self-consistent and avoids redundancy. It is also AFAIK much clearer than the current set of posted MAP document.

 An example of already achieved convergence is a proposal made by Ole.
 (It concerns the "BR rewrite fragmentation" item of his table of www.ietf.org/mail-archive/web/softwires/current/msg03364, i.e. disambiguation of datagram identifications in packets that go from shared-address CEs to the Internet.)   
 - The need to do something has been identified for the first time in the 4rd-U-03 draft.
 - Ole's proposed solution is simpler than that of the draft. (The need is satisfied directly in CEs, rather than left for BRs to satisfy it.)
 Unless a problem would be identified in the mean time, I will adopt Ole's solution in the next 4rd-U version. 

 Naturally, if Ole would agree to it, he would be most welcome as co-author of the upgraded unified specification. 


Best regards,
RD