[Softwires] ds-lite tunnel option

[Tomasz Mrugalski <[e-mail]>]

Hi,


>From what I understood during today's meeting discussion, there are 2 goals
that we are trying to achieve:


- Standarized option is urgently needed, so we should not prolong the
discussion more than necessary

- Generic purpose tunneling and encapsulation approach was discussed before
without any noticeable agreements


We have basically 2 proposals (Another proposed option mentioned in
draft-townsley-ipv6-rd6-00 is designed for DHCPv4, thus not applicable for
this discussion):

   1. Dual-Stack Lite option, proposed in
   draft-dhankings-softwire-tunnel-option-03. This is a simple solution,
   designed specifically for addressing dual-stack lite tunnel configuration
   (IPv4-over-IPv6). There are 2 independent implementations of that proposal
   available (ISC and dibbler, both open source).
   2. SC Discovery option, proposed in draft-guo-softwire-sc-discovery-01.
   The second proposal attempts to address generic tunnel configuration, over
   IPv4 and IPv6 as well. Two options (one for DHCPv4 and one for DHCPv6). Its
   generic approach comes at a cost. With 3 different fields for type
   specification and preference, the total number of fields is 7, plus
   additional possible sub options. Although 3 type fields (SC type, Tunnel
   Type and Protocol Type) give great flexibility, significant part of all
   possible combinanations is irrelevent or plain wrong. Therefore number of
   actual legal combinations is quite limited. Also, there is no
   implementation.

There is also certain confusion regarding the sc-discovery draft. -00 was
presented during meeting, but there is -01 version available. The only
noticeable change is Protocol Type field being extended to 16 bits and
ETHER-TYPE values are used.



Taking into consideration that there was a comment by an Internet Area
Director that DHCPv4 options specification is no longer a priority, I
propose to continue working on Dual-Stack Lite option, but to use some ideas
from SC Discovery option. As dual-stack lite is a IPv4/IPv6 only, there is
no need to specify tunnel type. Also, this will avoid confusion for vendors
if there are multiple encapsulations required. No, there is just single one:
IPv4/IPv6 only.


Should more generic tunneling configuration mechanism become required,
additional suboptions may be defined in the future.



What are group's thoughts regarding that matter?



If there is a contributting co-author position available, I would be happy
to volunteer.



Regards,

Tomek