On Dec 18, 2008, at 4:20 PM, Adrian Farrel wrote:

Hi Robert,

Thanks for reading.

The security consideration section is pretty light (particularly given how long this set of documents appears to have been cooking). Are there really no additional considerations over what's in [PCEP] introduced by this document? A couple of sentences further supporting that (if it's the case) would go a long way.

Sure. I think it is probably the length of cooking that led to the brevity of this section.

I guess if we added further text it would look something like...

PCEP security mechanisms are described in [PCEP] and are used to secure entire PCEP messages. Nothing in this document changes the message flows or introduces any new messages. This document introduces a single new object that may optionally be carried on PCEP messages and will be automatically secured using the mechanims described in [PCEP]. If a PCEP message is vulnerable to attack, for example because security mechanisms are not used, then the OF object could be used as part of an attack, however, it is likely that other objects will provide far more significant ways of attacking a PCE or PCC in this case.

We probably left this out because it was "obvious". Would it address you concerns if we added it?

Yes.



This is probably a naive question, but what happens if a PCC sets the OF flag in a request, doesn't provide a OF object, and the OF object in the response has a function code the client has never heard of before? (If this is clear, where in the document suite is it captured?)

The PCC is not required to process the OF object on the PCRep.
If the PCC gets a function code it doesn't understand it can:
- ignore it and use the response
- add the function code to its view of the PCE's repetoire for
later request/exclusion
- discard the response and ask another PCE to compute for it
- discard the response and ask the same PCE to compute
 excluding the function used.

I'm not sure we need to tell implementers how to write code here.

No, but making it clear to an implementer that this (and the following) are cases they will to have to deal with would be a good thing. Is the list you present above explicit (even if distributed) in other documents? - I'm definitely not well-versed in this protocol. But I'm more interested in making sure it's clear to implementors that the conditions can arise than laying out what they have to do about it - the
conditions are subtle if you read this document alone.



Also (and to check my understanding of how extensibility of the flags field in the RP object is supposed to work) - it's possible for the OF flag to be set in a request but for the response to not contain an OF object (if the PCE doesn't know about the extension in this draft it MUST ignore the non-zero flag bit according to the definitions on page 30 of [PCEP]). Should this document explicitly mention this case?

Right, this would be the behavior of a legacy node.
The PCC that set the OF flag would receive a PCRep carrying no OF object (i.e., in breach of this I-D).
The PCC would either:
- discard the PCRep because it *really* wanted the OF object returned
- shrug its shoulders and continue

Again, I'm not sure that we need to tell people how to write code. This certainly doesn't change the protocol behavior onthe wire.
see above


This draft uses RBNF (draft-farrel-rtg-common-bnf). The proto writeup indicates that no tool was used to verify the RBNF in the draft. Has there been a careful review by a human? (There isn't much of it, but enough for a typo or dangling reference to slip in).

Did you find any problems?

No, but I didn't try to either.

Spot-checking a couple of things:

Some productions use <SVEC-list> but the rule I think they're pointing to is <svec-list>. ([PCEP] is similar). Are the rule names supposed to be case-insensitive?

The rule you are updating from [PCEP] for PCReq has one of these: [<RRO>[<BANDWIDTH>]], which you are replacing with [<RRO>].
Is that intentional?



I wrote RBNF and also reviewed the I-D as doc shepherd. I found some issues in an earlier review and these have been fixed. I see no other issues.

||        Has the
||        Document Shepherd personally reviewed this version of the
||        document and, in particular, does he or she believe this
||        version is ready for forwarding to the IESG for publication?
|
|     Adrian Farrel is the document shepherd.
|     He has personally reviewed the I-D and believes it is ready for
|     forwarding to the IESG for publication.

I'm probably being dense, but I can't make the literal "0x200" in the definition of the OF flag make sense. It doesn't seem to have anywhere to go in the registry that [PCEP] creates. Additionally, that registry asks for a "Capability Description" and this document only provides a "Name" (of OF).

Yes, you're right. The IANA section of the PCEP spec has been updated several times in the last month and has left this I-D behind.

I think, based on Section 9.6 of [PCEP] we need to use Bit 25, but we need to check other I-Ds in the pipeline.

Section 8.2 is, if I read it correctly, a note that more standards work needs to happen, not that an implementer needs to do something. I suggest not using 2119 language here.

Yes, I'm happy with that.
I'm not sure the Ops ADs are so comfortable :-)

The rest of these comments are minor:

- The abstract does not capture that the document defines/registers objective functions/metric types

Add...

This document defines objective function code types for six objective functions previously listed in the PCE requirments work.
should the metrics also be called out?


- There's an extra closing parenthesis in the function on Objective Function Code 3: r(Lpi)) <-here

Good catch

- It would make IANA's job easier to point to the registries this document is adding values to (rather than creating) by the name. For instance, 6.2.3 of this document could say the "PCEP-ERROR Object registry" (assuming that's the name that [PCEP] is trying to get IANA to use for the registry it creates in its section 9.12. The [PCEP] document itself might benefit from being more explicit with registry names it suggests to IANA.

Well, that horse has bolted and we don't have names to point to.
So we will have to help IANA as the work goes forward.

Thanks for your review.
Adrian

_______________________________________________
Gen-art mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/gen-art

Reply via email to