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

Re: Multi6 WG Last Call (1 of 3) draft-ietf-multi6-architecture-01.txt



Sorry for not responding earlier; I was waiting for others' comments, and the ball got dropped. Inline..

On Wed, 3 Nov 2004, Geoff Huston wrote:
At 04:39 PM 3/11/2004, Pekka Savola wrote:
I've two bigger concerns here.  While they aren't changes which must be adopted, I think
doing so would improve the document's readability a lot and help in clarifying its scope, so it
might be worth the effort..

1)

The document is rather long and conversational in tone.


Yes. There is a fair amount of ground to cover and quite frankly in my case brevity would lead to a document that would quickly become cryptic. I chose this style as a) its a style I'm comfortable write in and b) I believe that it is suited to this form of architectural discussion, where the intent is not to be prescriptive in advocating one approach or another, but to look at the overall space and look at the manner in which various approaches choose to operate within this space.

Understood and agreed. Better be verbose than overly brief.

However, to be frank, I've had problems when reading the document. I've done it three times along the process, with different amount of concetration, and always I've thought, "so, what *were* the architectural issues?". What I want to say is that the document structure, subsection titling or some other details are probably interfering with the readability of the document.

That's probably not a big problem, but I found it difficult to read, because there was not a clear structure in the document. In particular, section 4.3 "Multi-homing: Identity Considerations" (which was discussing generic issues for further subsections) seemed ill placed to section 4. More appropriate place for quite a bit of this discussion would be under section 5, possible under 5.3.2 in particular.

I'm afaid that I do not agree with this. The purpose of this section is to motivate the following 2 section (4.3 and 4.5) by looking at the common aspects of identity in the context of multi-homed environments without assuming a particular implementation class.


Section 5 is a more general discussion about identities, and section 4.3 would be out of place if moved there.

Then maybe some other reorganization would be possible, or a change of focus. I'd just like to keep the sections concise, dealing with one particular issue only, so that it's easier to see the big picture without having to "combine the puzzle" in your head :-)


I'd like to reduce the amount of generic discussion in section 4, just
making it lay out the different approarches in sufficient detail, and
refer to the following sections for lengthier (and more structured)
discussion of the details.

This section (4) looks at generic styles of approach, rather than particular approaches, and looks at the potential advantages and possible limitations of multi-homing within the context of each style, and discusses them. The following section (5) is a more focussed examination of the behaviour and properties of identifiers.

As above, then maybe some generic properties of e.g. locator selection should be discussed in a separate section, appropriately titled subsections or the like.


2)

I've griped before about this document not having enough 'truth in advertising', i.e., it says it describes the architectural approaches to multihoming in IPv6, but does not go in sufficient detail to e.g., traffic engineering possibilities (the mention of this has been sprinkled along the draft though). Still, I think this is a bit of misleading and we should try to avoid this. I'd just make it clearer that this document focuses on connection survivability and not the whole multihoming problem space.

The document attempts to steer a neutral line between what some folk see as possibilities, and others see as unrealistic. For example, some would see host-based locator selection as a tool to push traffic along particular network paths which in turn could provide particular traffic engineering outcomes. while others would say that traffic engineering as an outcome of host-based locator selection is a case of fanciful thinking, as there is clearly insufficient network-based path state information provided to hosts to allow them to make an informed decision about a locator selection that would yield particular path characteristics and network load profiles.

Yes, some of these issues could be pushed through hosts and locator selection to the hosts, maybe using some tools by the network management. But if it's visioned this could be a significant part of the current solution set, then it should by all rights be expanded in the document to consider the different problems etc.


mostly editorial
----------------

  In addition, [2] documents further considerations for IPv6
  multi-homing.  Again, the reader is referred to this document for the
  detailed enumeration of these considerations.  The general topic
  areas considered in this study include:
  o  interaction with routing systems,
  o  aspects of a split between end-point-identifier and forwarding
     locator,
  o  changes to packets on the wire, and
  o  the interaction between names, endpoints and the DNS.

==> does this call for minor edits to update to the latest think-about-...
organization?

I do not believe so - what did you have in mind?

The document was just reorganized, while the topics are still the same, they should probably be classified here the same way as in the think-about.


  The possible application of the MIPv6 protocol to the multihoming
  problem would be to use BU messages to convey information about
  alternative addresses to be used after the outage.

==> clarify: s/convey/convey prior to the outage/ ?

An alternative wording is:

   The possible application of the MIPv6 protocol to the multihoming
   problem would be to use BU messages to convey information in
   advance about alternative addresses that could be used following
   an outage in the path associated with the currently used address.

OK

--
Pekka Savola                 "You each name yourselves king, yet the
Netcore Oy                    kingdom bleeds."
Systems. Networks. Security. -- George R.R. Martin: A Clash of Kings