Re: [trill] draft-ietf-trill-directory-assist-mechanism-10.txt

[Donald Eastlake <[e-mail]>]

Hi Sue,

Thanks for the review.

On Sun, Jan 8, 2017 at 6:02 PM, Susan Hares <[e-mail]> wrote:
> Donald, Yizhou, Linda, and Radia:
>
> Thank you for your work on draft-ietf-trill-directory-assist-mechanisms.
> Each draft improves the readability and nails down some of the potential
> edge cases.   I believe we are nearing the end of this journey toward IESG
> approval. I am joining with you on this journey by doing a shepherd review
> at the end of IETF LC.
>
>
> Status: Ready to publication, with some editorial nits you should consider
>
>
>
> Here’s my editorial nits on the latest version.   Please address #2 – that
> deals with the “SEND” functionality and #6
>
>
> #1, p. 5 paragraph 1, list item 3.   – simple spelling error
>
> Old:/MAC addresses “nomrally” /
> New: /MAC Addresses normally/

OK.

> #2. p. 17-22 – section 3
>
> #2.a Overall – I think you need to include SEND messages in this Query
> messages.

OK.

> #2.b p17, section 3.0 paragraph 5  specific text change
>
> Old:/The requests to Pull Directory servers are typically derived from
> ingressed ARP [RFC826], ND [RFC4861],  RARP [RFC903], or
> data frames with unknown unicast destination MAC addresses, intercepted by
> an ingress TRILL switch as described in Section 1.1/
>
> New:/ The requests to Pull Directory servers are typically derived from ARP
> [RFC826], ND [RFC4861], RARP [RFC903], or SEND [RFC3971] messages or data
> Layer 2 frames with unknown unicast destination MAC addresses intercepted by
> an ingress TRILL switch as described in Section 1.1/
>
> Reason: SEND mechanisms need to be clearly specified in the draft.  Suresh
> Krishnan mentioned this on the ARP optimization draft.

OK.

> #3 p. 19, section 3.2.1 paragraph 1, sentence 3
>
> Old: /The priority of the channel message is a mapping of the
>    priority of the frame being ingressed that caused the query with the
>    default mapping depending, per Data Label, on the strategy (see
>    Section 4) or a configured priority (DirGenQPriority, Section 3.9)
>    for generated queries./
>
> New:/ The priority of the channel message is a mapping of the
>    priority of the ingress frame which caused the query combined with the
>    default mapping per Data Label depending on the strategy for generated
>    queries (see Section 4) or a configured priority (DirGenQPriority,
> Section 3.9/.
>
> Reason: Clarify sentences.

I agree this could use clarification. How about:

   The priority of the channel message is a mapping
   of the priority of the ingressed frame that caused the query.
   The default mapping depends, per Data Label, on the strategy
   (see Section 4) or a configured priority (DirGenQPriority, Section
   3.9) for generated queries.

> #4, p. 33 section 3.5.1 paragraph 3
>
> Old:/ The Bridge shown might be a complex bridged LAN or might be absent
>    if, as shown for End Station 1, End Station 2 was dual ported with
>    point-to-point Ethernet links to RB1 and RB2./
>
> New:/The Bridge shown might be a complex bridged LAN, a LAN without a bridge
> (as shown in End station 1), or connected via point-to-point links (as shown in
> End Station 2’s which is connected through a bridge with point-to-point
> Ethernet links to
> RB1 and RB2./
>
> Reason: Clarify the sentences

OK.

> #5, p. 33 section 3.5.1 paragraph 4 sentence 1.
>
> Old:./
>    Because an indirect Pull Directory server discards information it has
>    cached from queries to an end station Pull Directory server if it
>    loses adjacency to that server (Section 3.7), if it knowns that such
>    information may be cached at RBridge clients and has no other source
>    for the information, it MUST send Update Messages to those clients
>    withdrawing the information/
>
> New: /Since an indirect Pull Directory server discards information it has
>      cached from queries to an end station Pull Directory server if it
>      loses adjacency to the server (Section 3.7), if it detects that
>      information may be cached at RBridge clients and has no other source
>      for the information, it MUST send Update Messages to those clients
>      withdrawing the information/
>
> Words changed – in bold
>
> Why: anthropomorphism – TRILL switches do not know.  TRILL switches detect
> based on logic.
>
> Why: Because/since – both indicate causes, but “since” seems to indicate an
> ordering that this paragraph suggests.

OK.

> #6: p. 43, 4th paragraph beginning with Although some …
>
> Current text:
>    /Although some of the ports sending TRILL ES-IS PDUs are on end
>    stations and thus not on routers (TRILL switches), they nevertheless
>    may make use of the Router Capability (#242) and MT-Capability (#222)
>    IS-IS TLVs to indicate capabilities as elsewhere specified./
>
> It would be good to indicate where these capabilities are specified.

OK, we can add a reference to RFC 7176

> #7 p. 44, section 6 – Security considerations
>
> After the SEC-DIR review comes in, consider if the end-system engage
> requires some extra text on privacy related to the end-systems.
>
> Since Donald and Radia are much better at all types of security, please
> consider this as a “please check”.   Kathleen and Stephen are focused on
> this work.

We could add something about edge RBridges being more aware of what IP
addresses are being used.

Thanks,
Donald
===============================
 Donald E. Eastlake 3rd   [phone] (cell)
 155 Beaver Street, Milford, MA 01757 USA
 [e-mail]