Nothing more from me (although I don't think that's where you query was directed), all the changes look good.

On the IANA question, was simply a matter of if you want to require something more rigid v. the most lax codepoint allocation model, but I don't have any strong opinion there.


On 2016-09-19 12:06 PM, Dino Farinacci wrote:
Any status on this?


On Sep 5, 2016, at 10:07 AM, Dino Farinacci <> wrote:

See responses inline. A new draft is attached (but not submitted since I had some open questions).

Document: draft-ietf-lisp-crypto-06.txt

Reviewer: Danny McPherson

Review Date: August 24, 2016

Intended Status: Experimental

Thanks for the review Danny.


I have some minor concerns about this document that should be considered before publication.


I believe the draft is technically sound.

Major Issues:

I have no “Major” issues with this I-D.

Good to hear.

Minor Issues:

In the Security Considerations section a small amount of text might be useful that discusses end-end v. encryption from middle boxes, and the risks therein. There are clearly benefits to this over no encryption, but there are risks about assumptions that may be made thereafter as well.

I added some text. Please review.


S.1: s/typically not modified. Which means/typically not modified, which means/


S.1: Is there in fact a case where asymmetries result in the *same* egress xTRs but different keys are used? I believe this would just apply to "different xTRs", no? :

Right, when the same ETR is used by different ITRs to encapsulate traffic, different keys are used.

However, return traffic uses the same procedures but with different key values by the same xTRs or potentially different xTRs when the paths between LISP sites are asymmetric.

Right, a unique set of keys are used for each {ITR, ETR} combination. Did you want me to say something specific here?

S.1: Regarding "[t]his document has the following requirements for the solutions space", it might be useful to reference some general IETF privacy work, even RFC 6973 or the like. Given that it's Experimental I think it's fine as is, but some references for the broader community may be in order. In particular, references to not requiring a separate PKI (and therefore external or circular dependencies!), avoiding third party trust anchor, rekeying as good operational practice, not just compromises, and other such arguments might be reinforced.

I added a reference to 6973.

S.3: Could include LCAF here, perhaps.

Added plus its reference.

S.4: You could probably strike this entire sentence and lessen confusion: "When an ETR (when it is also an ITR) encapsulates packets to this ITR (when it is also an ETR), a separate key exchange and shared-secret computation is performed.”

Okay, removed.

S.7: It’s unclear what constitutes “Diffie-Hellman *group*”.

I will change “group” to “parameters”.

S.7: s/the the/the/

S.7: s/integrity-check/integrity check/

Changed both.

S.8: Editors note to strike text in last paragraph here, unclear what resolution was from this text.

We allocated two new bits from a 3-bit reserved field now leaving 1 reserve bit. I added that the reserved bit remaining is documented in RFC6830.

S.12.1: A reference to the SAAG comments might be useful here?

We don’t have one. But I will add a list of recommendations they provided.

S 13: Are you sure you want a default FCFS allocation policy here and not a slightly higher bar?

I have no opinion. What is the benefit on not doing FCFS?

Throughout: Consistent hyphenation in the document would help (e.g., “network-byte” ..).


Throughout: Expanding on first use of each acronym would be useful, perhaps with references.


A new revision is attached (plus a htmlized diff file). Let me know if you like the text and then I’ll submit -07.

Thanks again,
Dino & Brian


lisp mailing list

Reply via email to