Adrian Farrel wrote:
Alexey, hello.
Thanks for the review.
In line.
Adrian
Minor issues:
1.1. Design Considerations
As is described in [RFC4379], to avoid potential Denial of Service
attacks, it is RECOMMENDED to regulate the LSP Ping traffic passed to
the control plane. A rate limiter should be applied to the
well-known UDP port defined for use by LSP Ping traffic.
What is this port? Is mentioning of the port significant?
It is well-known, so you obviously don't need to ask :-)
The port number is in the IANA registry and can be seen in RFC4379 (the base
definition of LSP Ping) sections 4.3, 4.4, 6, and 7.
The port number is 3503
I don't think it needs to be mentioned here. In general, restating definitions
causes potential conflicts. It is far better to have a normative reference to
RFC 4379 for all elements of LSP Ping that are not changed by this document.
Ok.
3.1.2.1. Multicast LDP FEC Stack Sub-TLVs
Address Family
Two octet quantity containing a value from ADDRESS FAMILY NUMBERS
in [IANA-PORT] that encodes the address family for the Root LSR
Address.
[IANA-PORT] IANA Assigned Port Numbers, http://www.iana.org
Which IANA registry do you have in mind? Seeing a link would be helpful.
Yes. This is a typo.
We fixed it for you in -17
Yes, I've noticed that.
3.2. Limiting the Scope of Responses
The P2MP Responder Identifier TLV only has meaning on an echo request
message. If present on an echo response message, it SHOULD be
ignored.
Are there known reasons for violating the SHOULD? I.e. what are the reasons
for having multiple sub-TLVs in the first place?
Future extensibility.
There is a corner of the world that builds conformance testers and would
interpret a MUST statement here as preventing any extensions that used this TLV
on an echo response as being in violation of this spec. Therefore we would have
to respin this spec if/when we made such an extension.
Furthermore, it is harmless to this protocol to not ignore such a TLV.
Ok. Being more explicit about this in the draft would be better, of course.
Anyway, in response to Dave Harrington's Discuss, we will be re-examining all
uses of "SHOULD" with a view to tightening them or supplying an appropriate
"MAY" clause.
3.2.2. Node Address P2MP Responder Identifier Sub-TLVs
The address in this Sub-TLV SHOULD be of any transit, branch, bud or
egress node for that P2MP LSP.
Is the use of SHOULD correct here (instead of a MUST)? Are there any choices
left if the SHOULD is violated?
The MAY clause can be added.
I think you've missed my point. If the list of node types is complete,
then there is no need to use a SHOULD (because it is a MUST and that
MUST doesn't need to be stated). Are there other types of nodes?
3.5. Downstream Detailed Mapping TLV
Downstream Detailed Mapping TLV is described in [DDMT]. A transit,
branch or bud node can use the Downstream Detailed Mapping TLV to
return multiple Return Codes for different downstream paths. This
functionality can not be achieved via the Downstream Mapping TLV.
Are "Downstream Detailed Mapping TLV" and "Downstream Mapping TLV" two
different things?
As
per Section 4.3 of [DDMT], the Downstream Mapping TLV as described in
[RFC4379] is being deprecated.
They are different precisely per the referenced text.
But well done! You caught a bug :-)
s/4.3/3.4/
4.1.2. Jittered Responses to Echo Requests
Echo response jittering SHOULD be used for P2MP LSPs. If the Echo
Jitter TLV is present in an echo request for any other type of LSPs,
the responding egress MAY apply the jitter behavior as described
here.
Can you provide a bit more information about how this work?
Obviously we authors think there is nothing more to say (or we would have said
it).
Maybe you can ask a more specific question?
If a jitter TLV shows up on an echo request for (e.g.) a P2P LSP, the egress may
delay sending the echo response in exactly the same way as the leaf of a P2MP
LSP would as described here.
Ok.
4.2.1.1. Responses from Transit and Branch Nodes
The presence of a Downstream Detailed Mapping TLV will influence the
choice of Return Code. As per [DDMT], the Return Code in the echo
response header MAY be set to value TBD ('See DDM TLV for Return Code
Am I correct that the value TBD is specified in [DDMT]?
You are correct as in the very next line that you cut off with your question.
I think it would be a good idea to insert a note for RFC Editor here
asking to fix TBD when [DDMT] is published. That wasn't very clear to me.
[...]
7.2. New TLVs
P2MP Responder Identifier TLV (see Section 3.2) is a mandatory
What does "mandatory" means in this section? Mandatory for IANA?
No. There are two TLV types. Mandatory TLVs and, erm, not mandatory ones. The
registry is split accordingly.
Ah, Ok.
TLV. Suggested value 11.
Four sub-TLVs are defined.
- Type 1: IPv4 Egress Address P2MP Responder Identifier
- Type 2: IPv6 Egress Address P2MP Responder Identifier
- Type 3: IPv4 Node Address P2MP Responder Identifier
- Type 4: IPv6 Node Address P2MP Responder Identifier
Echo Jitter TLV (see Section 3.3) is a mandatory TLV. Suggested
As above.
Also as above.
_______________________________________________
Gen-art mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/gen-art
