Re: [v6ops] draft-ietf-v6ops-6204bis WGLC

[Francis Dupont <[e-mail]>]

 In your previous mail you wrote:

>  Ole, authors of pcp-dslite, any SP testing PCP with DS-Lite, and v6ops,

=> seems I belong to this list (:-)

>  Please see below.
>  
>  Further draft-dupont-pcp-dslite-01 is not in RFC form nor in the IESG.

=> I don't parse the 'in RFC form': do you have some concerns about
the pcp-dslite document (other than below)?

>  Additionally the pcp-dslite document needs a peer review.

=> of course it does. Are you candidate to review it (or co-author it)?
BTW I can explain the pcp-dslite document history: there are two
possible different modes so I asked one and only one would be
chosen. The answer was "help yourself" so I polled other PCP implementors
to get their opinion and co-authoring, became the editor of the document
which in fact is mainly about the mode choice. Since the last version
the WG moved all the DS-Lite related details from the base spec to
the pcp-dslite document (so a new version will be needed, for instance
for some security considerations which was removed).

> What is the Appendix A in pcp-dslite talking about including mention
> of a "tag"?  Why is the tag needed when pcp-base already supports
> the necessary protocol?  Here is how.

=> it is an appendix (so not normative) which tries to explain how
to implement the "other" mode on the server side. BTW I used the verb
"tag" because it is less overloaded than "label" and stronger than "link".

>  The basic PCP header includes the source IP address of the PCP client
>  issuing the PCP request.  Thus even though the IPv4 PCP request reaches
>  the CGN in an IPv6 tunnel, the CGN decapsulates the tunneled packet and
>  passes the packet to the IPv4 stack on the CGN.  The IPv4 stack passes
>  the PCP Request to the PCP server.  Snipped from section 6.1 of pcp-base
>  it this text.
>  
>  [IPv4 is represented using an IPv4-mapped IPv6 address.]
>  
>  Thus the PCP server is able to reply back to the PCP client in an IPv6
>  tunneled PCP response back.

=> no, it is not able if it doesn't get the IPv6 address to select the
right tunnel, and this IPv6 is thrown away when the request tunneled is
decapsulated.
To summary you didn't understand the appendix A...

>  I question why pcp-dslite document is needed if pcp-base already covers
>  the technical details.

=> it is needed because there is an alternative to the encapsulation mode
(so two modes, i.e., one too many) which has some technical advantages
(mainly simpler to implement at the server side).

Note if you consider DS-Lite as being a particular case it should always
be a good idea to push it to a dedicated document as the base spec has
far too many things in it.

Regards

[e-mail]