Hi Martin
Sorry for the delayed response. Please find replies to your comments below.
Best regards
Jon
Could you elaborate on what you mean by: "or to perform a specific service on
the packet."
[Jon] I think it should say "specific function". Some examples of this can be
found in draft-filsfils-spring-srv6-network-programming. I think this is
fairly tangential to the PCEP document, so is probably not worth referencing,
and perhaps we should remove the text about "specific service" altogether.
s/A Segment Routed path/A Segment Routing path/
In SR networks, an ingress node of an SR path prepends an SR header to
all outgoing packets. The SR header consists of a list of SIDs (or
MPLS labels in the context of this document).
This appears twice in two consecutive paragraphs at the beginning of section 3.
[Jon] Thanks - the redundant text will be deleted.
I'm not a native english speaker but I find the use of "ERO subobject"
confusing when used to refer to the SR-ERO subobject. Maybe say the "the
(SR-ERO) subobject of the ERO".
[Jon] I think the original is OK. In general, if I read "the <foo> subobject",
I interpret that as "the subobject of type <foo>".
Length: Contains the total length of the subobject in octets,
including the L, Type and Length fields. The Length MUST be at
least 8, and MUST be a multiple of 4. An SR-ERO subobject MUST
contain at least one of a SID or an NAI. The length should
include the SID and NAI fields if and only if they are not absent.
I am confused about the minimum length. L, Type and Length use 2 octets. What
bring this to 8 considering that a SID is 4?
[Jon] The NT and Flags fields add another 2 octets. The diagram shows them but
the text does not mention them, whereas it does mention L, Type and Length.
Perhaps this would be less confusing:
OLD
Length: Contains the total length of the subobject in octets,
including the L, Type and Length fields.
NEW
Length: Contains the total length of the subobject in octets.
END
Also the last sentence is very confusing. It seems normal not to count
something which is not there. No need to say so. The current statement seems to
contradict the preceding sentence in fact, so I'd just remove it.
[Jon] Agreed.
* A 4 octet MPLS label, where the 20 most significant bits encode
the label value per [RFC3032].
s/MPLS label/MPLS Label Stack Entry/
I believe it is too late to change but I find L=1 meaning "no limit" is *very*
confusing. For me L stands for Limit and when L=1 there is a limit, when L=0
there is none.
[Jon] Agree, both that it is confusing and too late to change :-)
On the metric object.
You have the requirement that "the PCE MUST minimize the SID depth". This is a
non actionable requirement. You need to set a bound. I very much understand
that the bound is the specified sid depth (and you say it properly afterwards
("the PCE MUST NOT return a path whose SID depth exceeds the given
metric-value"), but still the sentence I point to is a non actionable
requirement which you may want to reformulate.
[Jon] Agreed. How about "the PCE minimizes the SID depth".
If a PCEP session is established with a non-zero default MSD value,
then the PCC MUST NOT send an MSD METRIC object with an MSD greater
than the session's default MSD.
Out of curiosity, why do you forbid that case?
[Jon] The PCC starts a session saying its MSD is 4. It then requests a path
saying that the MSD must not exceed 5. This is a clear contradiction.
_______________________________________________
Pce mailing list
Pce@ietf.org
https://www.ietf.org/mailman/listinfo/pce
