On Tue, Mar 01, 2011 at 09:08:51PM +0000, Lee, Yiu wrote:
> [YL] I agree with you this have to be specified. I can think of either the
> cpe-router-bis draft or the draft-dslite-deployment should specify the
> exact behavior. Do you agree that the draft-dslite-deployment is a good
> place to define this behavior?

Not really. Operators need a normative specification to point CPE router
vendors at. draft-dslite-deployment aims to be "Informational" and
"discusses" things. It doesn't mandate specific behaviour. The strongest
guidance I read in the document is that B4 SHOULD implement the DHCP
option.

It's already hard to push vendors to implement things in moving drafts,
but it will be even harder to ask them to implement what they read into
informational documents. :-)

On a side note, it would be reallyreally good if we would get the DHCP
option draft to RFC as quick as possible in order to have vendors
actually implement the proper final IANA assigned DHCP option -
currently it's all custom handwaving, making interop testing hard at
best.

> With the open-source code I use for test, I have to manually configure the
> CPE to the dslite mode, but this is less than ideal.

Indeed. Current implementation we test with has a GUI switch:

(*) native IPv4 (DHCPv4)
( ) DS-Lite

No "auto" heuristics, and DHCPv4 being default with no automatic DS-Lite
"fallback". In fact, any customer with a DS-Lite-only provisioned
connection would have to change that option in order to come online
proper if it would remain like that.

Next week we'll get the next DS-Lite pre-production test implementation
of another major CPE router vendor on the test bench - I'm eager to see
how DS-Lite "optionality" is done there. They are expecting us to tell
them how it's done "right" from an operational point of view - but
that shouldn't be ISP-specific, but well reviewed guidance from
some normative spec (sorry, I'm repeating myself I guess).

> I remember I previously discussed this with the cpe-router-bis's
> authors. I will have to dig up the old mails to find out what
> conclusion we reached.

Thanks!

Best regards,
Daniel

-- 
CLUE-RIPE -- Jabber: [email protected] -- dr@IRCnet -- PGP: 0xA85C8AA0
_______________________________________________
Softwires mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/softwires

Reply via email to