Excerpts from Brian E Carpenter on Wed, Feb 18, 2009 01:58:26PM +1300:
> Probably useful to state that this is layer 4. I understand that over
> at the ITU-T they have changed the meaning of the word 'transport' to
> mean layer 2 (presumably in an effort to forget about OSI).

They forgot about OSI years ago.  For a good time see
http://www.itu.int/rec/T-REC-Y.110/en on roles and "added value",
http://www.niven-jenkins.co.uk/An%20introduction%20to%20functional%20architecture.pdf,
and http://www.itu.int/rec/T-REC-Y.2611/en for functional
architecture.  For example in Y.2611 (where "identifier" is a
more-generic including both "name" and "address"):

    Identification is required in each layer network of the NGN
    transport stratum. A given entity will be assigned one or more
    identifiers depending on the function of that entity. FPBN layer
    network identifiers are independent of any client (and any server)
    layer network identifiers even if they share the same syntax or
    structure. At the boundary of a layer network, mapping and/or
    translation mechanisms are required in order to set up
    relationships between the identifier used by the client layer
    network and the identifier used by the server layer network.  

    NOTE – Identifiers could be determined from multiple discontinuous
    fields. The global uniqueness of an identifier may be provided by
    the context as well as the identifier itself. 

    Whether a given identifier is considered to be a name or an
    address is dependent on several factors including the perspective
    (and location) of the entity that is using (or mapping to) that
    identifier.  The same identifier can be considered an address to
    one entity and a name to a different entity because their
    perspective is different. 

> >    The consensus of the
> >    group is that such site renumbering is a completely unacceptable
> >    requirement and as such, these types of solutions are not of interest
> >    for further exploration.
> 
> Mild rewrite as
> 
>    The consensus of the
>    group is that such site renumbering is widely unacceptable for
>    operational reasons and thus, these types of solutions are not of
>    interest for further exploration in this group.

+1

> > 3.1.2.  Translation
> 
> Reading further, it occurred to me that this really should
> be called Map & Translate

I don't see why.  Suppose I'm doing GSE.  I have a simple algorithm
for swapping the upper bits of an address field, and the algorithm is
always the same.  Is that really mapping?  I don't know of any
proposals that do actual mapping and then translating.

> , and that we should distinguish
> stateful and stateless translation.
> 
> Also distinguish reversible and non-reversible translation.

Yes

> It would be good if this and the following section brought out
> the isomorphisms and differences between translation and
> encapsulation very clearly.

Yes

> (NAT is just swapping RLOC and EID, for example, but NAPT is
> munging EIDs together before doing the swap.)

I don't know what you mean, munging EIDs together.  Do you mean
that it uses pieces from different headers?

Scott
_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg

Reply via email to