Re: comments on requirements-05

[Iljitsch van Beijnum <iljitsch@muada.com>]

On vrijdag, mei 9, 2003, at 19:23 Europe/Amsterdam, Joe Abley wrote:

> > This is completely confusing. It seems we lack a definition for 
> > "site", and as a non-native speaker I might be missing something, but 
> > the word "site" makes me think along the lines of a building site: a 
> > single location. Transit providers tend to operate networks that span 
> > more than a single location.
> The document includes a definition for "site":
Ok, I took Thierry Ernst's message as an indication there was no 
definition; my apologies.

>    A "site" is an entity autonomously operating a network using IP and,
>    in particular, determining the addressing plan and routing policy 
> for
>    that network. This definition is intended to be equivalent to
>    "enterprise" as defined in [2].
> > If we're going to change wording, let's do away with "site" and 
> > replace it with "AS". This one is well-defined and used extensively 
> > in discussions about interdomain routing.
> The original reason for not using "AS" was that we did not want to 
> presuppose a routing solution to the problem (e.g. to the exclusion of 
> solutions which involve address overloading on the edge).
Ok, we keep "site" then even though the definition seems identical to 
AS, but I do maintain that "transit site" is confusing and shouldn't be 
used.

> > >    A transit provider's site is directly connected to the sites for
> > >    which it provides transit.
> > Nonsense.
> As I seem to be mentioning in every single message on this subject, 
> the reason for the definitions being there at all is that different 
> people have different ideas about what "transit provider" means.
When I was responding everything was pretty much out of context. I took 
"sites for which it provides transit" to mean "destinations reachable 
through the transit service" which would indeed be nonsense, but the 
context makes clear (or at least: clearer) that this means "sites that 
are customers of the transit provider". Maybe it's a good idea to spell 
this out in the definition.

> At home, I buy transit from an ISP in Ontario called Sentex. They buy 
> transit from Telus. According to some understandings of "transit 
> provider", Sentex is a transit provider of Joe Abley, but Telus is 
> not, since Telus have no direct provider relationship with Joe Abley.
> According to other definitions, Telus and Sentex are both transit 
> providers of Joe Abley.
Regardless of your personal circumstances, they are always "a transit 
provider", which is the term being defined. Things would be different 
if you were defining "a site's transit provider".

> To clear up this ambiguity, the document was modified to include 
> definitions. To be clear, the purpose of the document is not to define 
> the term "transit provider": the purpose is to describe a set of 
> criteria which might be viewed as useful capabilities for a 
> multihoming strategy. If the set of definitions in the document are 
> (a) not completely off the wall and (b) cover the terms used in the 
> document, then we are finished talking about definitions.
Nobody is ever finished talking about definitions.   :-)

> > Ok, now define the internet...
> It's the "largest equivalence class in the reflexive transitive 
> symmetric closure of the relationship "can be reached by an IP packet 
> from" [Seth Breidbart]
That's a good one.

> > Logical connections work too.
> Let's not introduce new exciting vague terminology for the sake of it.
If I am connected to my ISP over ADSL, how is that a "physical" 
connection with the ADSL whole sale company's ATM network in the 
middle? If the connection isn't always physical, the word physical 
shouldn't be used.

Iljitsch

PS. I really do appreciate you doing a difficult job here.