Hi Benjamin,

Thanks for your comments. The following edits address most of them:

https://github.com/paololucente/draft-ietf-grow-bmp-adj-rib-out/commit/2134855587c0b98f24533aa66d76e05f3fa6ed29#diff-42c33450ab1d1f6d1ec35e9a12c7bf24

My comments inline:

On 9 Jul 2019, at 00:40, Benjamin Kaduk via Datatracker 
<[email protected]<mailto:[email protected]>> wrote:

Benjamin Kaduk has entered the following ballot position for
draft-ietf-grow-bmp-adj-rib-out-06: Discuss

When responding, please keep the subject line intact and reply to all
email addresses included in the To and CC lines. (Feel free to cut this
introductory paragraph, however.)


Please refer to https://www.ietf.org/iesg/statement/discuss-criteria.html
for more information about IESG DISCUSS and COMMENT positions.


The document, along with other ballot positions, can be found here:
https://datatracker.ietf.org/doc/draft-ietf-grow-bmp-adj-rib-out/



----------------------------------------------------------------------
DISCUSS:
----------------------------------------------------------------------

The "Peer Up message Information" TLV type seems under-specified and
under-motivated.  (It is not mentioned in Abstract or Introduction.)  Why
does it need to be defined in this document, and what role is it
expected to play?  Who is the expected audience for it?  Is it limited
to the "group name"-like functionality described in Section 7.1?  Why is
cleartext appropriate, and are there any potential privacy
considerations for any potential use cases?

I’d +1 the reply from Jeff on this.

----------------------------------------------------------------------
COMMENT:
----------------------------------------------------------------------

Section 1

                      An example of pre-policy verses post-policy is
  when an inbound policy applies attribute modification or filters.
  Pre-policy would contain information prior to the inbound policy
  changes or filters of data.  Post policy would convey the changed
  data or would not contain the filtered data.

Can applying policy ever act as injecting new data?

Yes. Jeff Haas proposed the edit below to draft-ietf-grow-bmp-loc-rib-04 — 
would the same/a similar edit address your point?

https://github.com/paololucente/draft-ietf-grow-bmp-loc-rib/blob/master/draft-ietf-grow-bmp-local-rib.txt#L682-L684

Separately, I'd ask whether it's worth mentioning the new statistics
types in the Introduction (skipping them from the Abstract is probably
understandable).

My feeling is that mentioning new statistics tyeps is too much detail for the 
introduction as well. While they integral part of the draft, they are not part 
of the core concept.

Section 4

  o  Peer Address: The remote IP address associated with the TCP
     session over which the encapsulated PDU is sent.

  o  Peer AS: The Autonomous System number of the peer from which the
     encapsulated PDU was sent.

  o  Peer BGP ID: The BGP Identifier of the peer from which the
     encapsulated PDU was sent.

I am not sure whether I'm reading these properly.  Backing up, in
regular RFC 7854 BMP, we have a BMP sender (router) that is speaking BGP
to (presumably many) peers.  These Peer Address/AS/BGP-ID fields are
attributes of the BGP connections the router has to its peers, and are
talking about how the data is coming in.  For the Adj-RIB-Out version
that we're defining in this document, we want to flip the sense, so that
we are talking about what the router is sending on its outgoing BGP
connections.  In this way, the content being sent over BMP still
reflects the origin of the BGP data, which makes sense.  What's tripping
me up is that we still talk about "the peer from which the encapsulated
PDU was sent" -- IIUC this is "peer" in the "BGP peering" sense, so the
data being sent over BMP is the local data from the router.  Do we need
to say "peer" at all, here, then?  So, "The AS number for which the
encapsulated PDU was sent", and "The BGP Identifier for which the
encapsulated PDU was sent”?

You are right. I catched this before in fact you see the “Peer Address” 
description actually makes sense; we forgot to edit accordingly the other two, 
Peer AS and Peer BGP ID. I did the edit now.

  o  Timestamp: The time when the encapsulated routes were advertised
     (one may also think of this as the time when they were installed
     in the Adj-RIB-Out), expressed in seconds and microseconds since
     midnight (zero hour), January 1, 1970 (UTC).  If zero, the time is
     unavailable.  Precision of the timestamp is implementation-
     dependent.

Are leap seconds included in this value?  (Yes, I see that this language
is basically taken directly from RFC 7854, which also left it
underspecified.)

I would propose to leave it underspecified as well, leaving it implementation 
specific.

Section 5.1

                                                              Some
  attributes are set when the BGP message is transmitted, such as next-
  hop.  Adj-RIB-Out Post-Policy MUST convey what is actually
  transmitted to the peer, next-hop and any attributes set during
  transmission should also be set and transmitted to the BMP receiver.

I'm having a bit of trouble matching up "MUST convey" with "should also
be set".  (Also, we seem to say "MUST be what is actually sent to the
peer" in Section 3; do we need the normative language in both places?)
nit: this is a comma splice

Ack, i removed the “should also be set” part. I have also normalised language 
among the two places. I feel it is good to stress the concept in the two 
places, despite the text repetition.

Section 5.2

  Some attributes are set only during transmission of the BGP message,
  i.e., Post-Policy.  It is common that next-hop may be null, loopback,
  or similar during this phase.  All mandatory attributes, such as

nit: I suggest clarifying that "this phase" is pre-policy, not "during
transmission”.

Ack, fixed.

Section 6.2

Do we need to say anything about the byte order of the 2- and 8-byte
fields in these statistics structures?

My take would be that it’s implied that bytes are transmitted as network byte 
order / big endian. rfc7854 does not specify this either. Nevertheless if it is 
felt a key detail, i can certainly add a remark about it.

Section 6.3

                         BMP receiver implementations SHOULD ignore the
  O flag in Peer Up and Down notifications.  BMP receiver
  implementations MUST use the per-peer header O flag in route
  monitoring and mirroring messages to identify if the message is for
  Adj-RIB-In or Adj-RIB-Out.

Is this last MUST duplicating content elsewhere in the document?  It
doesn't seem related to the title of the section, "peer down and up
notifications”.

Agree, removed.

Section 7.1

Do I interpret this correctly as saying that peer and update groups are
not a defined protocol feature but rather something offered by
implementations for convenience of administrators?  The language about
"should be simple to include a group name [...] but it is more complex
than that" leaves me a little confuesd about what is current deployed
reality, what is speculation about potential future work, and what is
being defined as an actual new protocol feature.

I’d +1 the reply from Jeff on this.

Section 8

The administrative information TLV has some considerations about what
kind of internal organizational information is shared to "the world".

As Alvaro notes, publishing both pre- and post-policy outbound RIBs can
give a new information channel into what the policy applied to outbound
routes is, and presumably the usual BMP configuration considerations
apply about only sending information to people authorized to receive it.

Ack, edited. Please see my reply to Alvaro.

Section 9.3

The registry where this seems to be listed claims to be the "BMP
Initiation Message TLVs" registry; is the section heading of "Peer Up
Information TLV" appropriate?

There is a draft undergoing to split Initiation Message and Peer Up message TLV 
registries, see https://tools.ietf.org/html/draft-scudder-grow-bmp-peer-up-00 , 
which makes a lot of sense and is being adopted by the WG. I look for 
suggestions on what would be the nicest way to address this.

Paolo

_______________________________________________
GROW mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/grow

Reply via email to