[CCWG-ACCT] What do we mean by "service"?

[Eric Brunner-Williams]

Colleagues,

What distinguishes our area of work from all others (that I can think of 
off the top of my head, and some five decades of programming, mostly 
distributed systems), is that the uniqueness of the identifiers (and we 
can ignore protocol parameters, which need not necessarily be "numbers", 
btw), both "names" and "numbers", doesn't take on any particular 
"uniqueness" meaning until the prefix for any particular name-to-address 
mappings are announced and can be referenced from "elsewhere".

So the instances of communication that concern us are those which the 
address (actually its prefix) of some device is "announced" (modernly an 
ASN published via BGP4) and is known, or knowable, to a device at some 
other address (actually some other prefix), similarly "announced".

Where these "announcements" of ASNs and the associated prefixes (via 
BGP4) are both known is in a table in some intermediate device, which 
knowing both ASNs is capable of forwarding packets from one ASN to 
another, with each ASN undertaking the "last hop" or "default" local 
routing to each of the two devices conducting some instance of 
communication.

We refer to these intermediate ASN-knowing devices as "routers", and 
those which "take a full (routing) table" (a) have a _lot_ of memory 
(see Note infra) and (b) form a system which, because announcements and 
withdrawals of prior announcements occur (mostly) asynchronously, is 
modernly best thought of as forming the "skitter core" (kudus to kc 
claffy and CAIDA for the original research and the circular diagrams 
(some found on the halls of ICANN during its Admiralty Way days)).

Back in the EGP period we could think of an "internet core" as existing 
in some deterministic sense. Modernly, what the global routing table 
looks like depends on where we are looking from, and our notions of 
"global routing tables" are statistical in nature.

Those of you coming from all important names side of the business may be 
slightly discomforted by the notion that names don't matter until they 
are routed, and routed through the Default Free Zone (DFZ), but that's 
what we really mean when we speak about "globally unique identifiers" -- 
not the maximal string spaces that can be formed by 63 bytes, nor all 
the IPv4 and IPv6 addresses, nor all possible parameters the IANA has 
been tasked to publish, but only those addresses who's prefixes are 
announced to the DFZ, and those strings associated with those addresses 
by some names registry.

So just as Signaling System 7 (defined by the ITU-T's Q.700-series 
recommendations) is what we really mean when we say "telephone", that 
is, a device capable of using the Public Switched Telecommunications 
Network allowing us to dial-in to a CCWG call and communicate with one 
another, resolution of a name across the DFZ by two (or more) devices is 
what we really mean when we, wearing our {ICANN|*NSO|ASO|*AC|DNSOP} 
policy hats, say "service" (smile optional).

This multiply interrupted modest tutorial for some of the people I know 
can be battered into a para, perhaps even a sentence, with clear 
meaning, as an alternative to the existing "service" and the alternate 
"e.g., email, ..." language of the current drafts, and it can be 
discarded as correct but not useful, or incorrect _and_ not useful. It 
may assist the final word smiths.


Eric Brunner-Wiliams
Eugene, Oregon

Note: As of 10/09/15 179,957 ASNs were announced in the ARIN region, 
another 143,717 ASNs were announced in the APNIC region, and another 
136,435 ASNs were announced in the RIPE region, another 60,747 ASNs were 
announced in the LACNIC region, and another 11,914 ASNs were announced 
in the AfriNIC region.