[trill] Shepherd's review of draft-ietf-trill-centralized-replication-05

Donald Eastlake <d3e3e3@gmail.com> Mon, 20 June 2016 22:15 UTC

MIME-Version: 1.0
From: Donald Eastlake <d3e3e3@gmail.com>
Date: Mon, 20 Jun 2016 18:15:02 -0400
Message-ID: <CAF4+nEFpoTGsupLb+OBmRSw7spdMY8p725WER26gT7mnsGdhew@mail.gmail.com>
To: "trill@ietf.org" <trill@ietf.org>
Content-Type: multipart/alternative; boundary="001a113e3a8c093e660535bd0aae"
Archived-At: <https://mailarchive.ietf.org/arch/msg/trill/c4NAUbDG0ohk8vvzTdftMnRZiIw>
Cc: "trill-chairs@ietf.org" <trill-chairs@ietf.org>
Subject: [trill] Shepherd's review of draft-ietf-trill-centralized-replication-05
Precedence: list

Hi,

As the document shepherd I have done a careful review of this document and
have the following comments:

This draft uses "CE" and expands it as "Classic Ethernet"; however, I
believe we have decided to use the expansion "Customer Equipment"
that, for example, is used in RFC 7379.

The draft says that BUM traffic is unicast encapsulated and is sent to
a central node that SHOULD be a tree root. It would be good to say a
few words about why it "SHOULD". Perhaps something like "because
distribution tree roots are normally chosen to be high capacity core
RBridges with many high bandwidth adjacencies".

At the bottom of page 5 there appears to be a small garble where some
text was dropped. Looking back at version -04, it looks like most of
the problem is one line of text being dropped.

In TRILL it is possible for an RBridge to hold multiple nicknames and
be the distribution tree root for multiple trees. The draft should say
something about what to do in that case. Presumably, to avoid
unnecessary re-ordering of BUM traffic each R-nickname should direct
traffic to one distribution tree. Also, a centralized node could hold
multiple nicknames that it advertised as R-nicknames so perhaps the
spreading over centralized nodes in Section 8 should be a spreading
over R-nicknames. This would enable a network manager to direct more
traffic to one centralized node and less to other centralized nodes
which they might want to do if one centralized node was larger and
more capable than others.

In Section 11, it says that the C-nickname is set on each edge
RBridges which could be a bit confusing. It sounds a bit like they are
setting it on their own nickname. Should this say something more like it is
set on the psuedo-nickname for each edge group.

I also have a number of minor editorial suggestions which I am sending to
the authors.

Thanks,
Donald
===============================
 Donald E. Eastlake 3rd   +1-508-333-2270 (cell)
 155 Beaver Street, Milford, MA 01757 USA
 d3e3e3@gmail.com

Re: [trill] Shepherd's review of draft-ietf-trill… Haoweiguo
[trill] Shepherd's review of draft-ietf-trill-cen… Donald Eastlake