[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: I-D ACTION:draft-baker-v6ops-end2end-00.txt


On Aug 16, 2005, at 9:27 AM, Eric Klein wrote:

Fred,

A few comments (mostly nits).

1. This is the first time I have seen a draft that refers to it's self as a
note ( in the abstract and introduction), this seems like the wrong word.

Without taking the time to do a full investigation, I note that RFC #10 opens with the phrase "This note is a revision of NWG/RFC #3", 8 RFCs (997, 1020, 1062, 1117, 1150, 1244, 1311, and 1543) refer to the RFC series as a "series of notes", that 623 RFCs contain the word "note" on the first page (but I didn't check all of them to see whether it was used as an adjective or as a verb), and that a related series was once entitled the Internet Engineering Notes, or IENs.

I dunno, what word would you suggest? Some documents refer to themselves as "memos" or "memoranda"... :^)

2. In the last sentence of section 2.1 it reads:
   "IPv4, there is at minimum IPv4 connectivity",
I recommend it read:
  "IPv4, there is at a minimum IPv4 connectivity"


OK

3. In the last sentence of section 3 it reads:
' "A satisfactory solution", it could be said, "contains no unnecessary
complexity", and "first, does no harm." '

This seems a little confusing, as the one with first is second or last. It
may be confusing for non-native English speakers.

I had rather hoped that these were obvious fractured quotes relating to widely-held principles. I'm not finding it on the web at this instant, but one of the various latin statements of Occam's Razor that I have run across translates as "a satisfactory proposition contains no unnecessary complexity". The Hippocratic Oath is often quoted as the source of "First, do no harm"; the words are in fact not found there, but might derive from a related work (also by Hippocrates) called "Epidemics", or from another work by Galen. Galen's words appear to have been "Primum non nocere".

If you would like, I can remove the word "first", or directly quote "Epedemics", which says "make a habit of two things—to help, or at least to do no harm".

(punching "Occam's razor" and "Hippocratic Oath do no harm" into google yields an instructive list of references, for those who are interested)

4. In section 3.1, in the third paragraph ("If the central IPv4 networks are
providing the"), is this referring to the backbone or interconnection points
(like the public MAE (east/west) Internet Exchange or any of the hundreds of
private ones)? Are there any drafts that talk about converting the backbones
or interconnection points to IPv6 using any of the methods used (tunnel
broker, etc) or are they all end network based (like Tredeo, etc). 

There are general comments on tunneling in:

http://www.ietf.org/rfc/rfc2893.txt
2893 Transition Mechanisms for IPv6 Hosts and Routers. R. Gilligan, E.
Nordmark. August 2000. (Format: TXT=62731 bytes) (Obsoletes RFC1933)
(Status: PROPOSED STANDARD)

and specific comments relating to ISP networks in:

http://www.ietf.org/rfc/rfc4029.txt
4029 Scenarios and Analysis for Introducing IPv6 into ISP Networks. M.
Lind, V. Ksinant, S. Park, A. Baudot, P. Savola. March 2005. (Format:
TXT=64388 bytes) (Status: INFORMATIONAL)

The recommendation of the latter (section 3.2.2) is that the ISP run a native dual stack internally (ie, route both IPv4 and IPv6 prefixes using appropriate routing protocols, and forward both IPv4 and IPv6 datagrams natively) and allow their customers to connect to that (section 3.2.3) using IPv4 initially and IPv6 later as they come up.

I can list a number of networks that have literally taken that approach. I can list a number of others that have operational concerns, and have chosen instead to run parallel IPv4 and IPv6 networks internally, and either sell the customer separate IPv4 and IPv6 connections or split the traffic into those networks from their edge router. Perhaps half a dozen that I know of mention running an IPv6-only core and tunneling their IPv4 traffic over it. I obviously can't list my customers here, but if a network wants to identify itself, this would be a good time to do so... I make the point in 2.1 that these networks don't bother me: within a single administration, they can make it operate correctly for themselves, their customers, and their peers.

What I *am* concerned about is the number of networks that are telling me that they want to run IPv6-only at this early stage in the transition, and the proposals being made for running IPv4 over those IPv6 networks. DSTM is one example; there are other approaches being discussed as well. In at least some cases, these are themselves networks of networks - a central IPv6-only core operated by one administration is tunneled by its customers (other administrations) to deliver IPv4 service among them. Here, I get very worried very quickly. While it will no doubt work well enough for those that are members of the club, interconnection with anyone else is not guaranteed. In other cases, I am being told by several networks that they expect to have limited IPv4 requirements and expect their supporting networks to make good things happen for them. However, since I talk with all of them, I know that they are supporting networks of each other, and can see that they aren't necessarily providing the supporting services to each other. Again, I get very worried very quickly.

I really want to hear that they all will run native IPv4 and IPv6 routing and forwarding at their edges (eg, running dual stack networks), and not expect their customers to tunnel them. I really and truly want to hear that.