ietf-smtp
[Top] [All Lists]

Re: ADMD

2007-03-14 11:05:13



David MacQuigg wrote:
The Border is a special ADMD boundary, one in which there is no prior relationship.

That strikes me as an extremely helpful definition. Concise, clear and meaningful.

I think, however, I might disagree with it, although I'd be inclined to put this in terms of a change, rather than being basic.

I believe that a Border MTA role is important, even when there is a special arrangement between the neighboring ADMDs. (It doesn't matter whether the IP linkage between the ADMDs is through a private network or over the public Internet. There are all sorts of special trust arrangements possible.

1. Do folks prefer "Border" or "Boundary"? I think "Boundary" gets used more, but I don't really care which is chosen, as long as there is good agreement to yse the term?

So I'd be inclined towards a definition along the lines of:

2. "A Bo* module is portal to an ADMD. It may be provide any of the underlying functional roles within the architecture. Its additional role at the Bo* is to enforce exit and/or entry policies for the ADMD, when interacting with Bo* modules in other ADMDs."

Thoughts?


You define the term "Edge" on page 12 to mean something similar, but Fig. 4 shows an Edge between two related ADMDs.

Mumble. The document uses the term "Edge" to refer to an entire ADMD rather than to a module within it. An Edge ADMD is an originator or final recipient administrative environment.


A Bo* module will exist at the outer limits of *any* ADMD, edge, transit, or whatever.

No they are not. They are part of an integrated service in which you participate. The relationship between you and the other provider is important, but it does not make them part of your network, except perhaps for a non-technical use of the word network.

I would think the term "network" should include a collection of agents connected by contractual relationships.

Although that is a perfectly reasonable definition, it is not the way the term is used in the computer science and IT profession of "networking". I think we should be very careful *not* to assign unusual meanings to terms of art within the profession.


Anyway to facilitate our discussion I will call what I have an NoA (Network of ADMDs). My NoA is what outsiders have to deal with when they exchange email with me.

Well, I certainly see some benefit in having a term that refers to an aggregation of ADMDs that interact, or have some relationships, or otherwise are worthy of being considered together.

What do other folk think about this?


I'm not saying that ADMDs aren't important, just that there is a higher level of organization, a Network of ADMDs that should not be ignored. When I first read this document two years ago, I came to the conclusion that IP-based authentication was futile, because a chain-of-trust through all these ADMDs was too difficult to establish. Now I see that there is a simpler way of looking at this architecture, one that facilitates IP-based authentication at a single, well-defined Border. The document is misleading because it implies that such a Border does not exist.

1. It isn't meant to have that implication.

2. I certainly agree that IP-based authentication has utility between directly-interacting sites. But then, I am an author of CSV... And in any vent, discussing this in detail is off-topic, although using as an exemplar seemed quite useful.

Can we at least agree that there is one special boundary between ADMDs, and that is the Border between sending and receiving NoAs?

I think it is between any two ADMDs. Not sure I understand the special role between "NoA"s. Please clarify.


 Can we agree
that this Border is important enough to be included in Figs 3, 4 and 5? Do you see that this Border is unique in that

Maybe.

As you already know, there is a serious challenge to keep the figures understandable.

d./
--

  Dave Crocker
  Brandenburg InternetWorking
  bbiw.net