[CCWG-ACCT] Attempt to summarize discussion regarding Mission and Contract

[Andrew Sullivan]

Hi,

I sent an alternative to this in the "whole text" message I just
mailed, but I thought it better to follow up to this specifically in
this thread.

On Thu, Nov 12, 2015 at 04:05:40PM -0500, Greg Shatan wrote:
>    - services (i.e., the software processes by which commands received via
>       the Internet are processed and a response is generated and
> transmitted via
>       the Internet, to be viewed in a web browser, email client, or the like)
>       which use the Internet’s unique identifiers, or

That can't be the definition of services, I think.  First, I'm not
entirely sure that we want to say "receives commands": DNS, for
instance, is a completely bilateral protocol that just sends messages
back and forth.  The messages are the same format in "command" (query)
and response, differeing technically only in the setting of a bit.  I
_could_ make an argument that these are commands, but I don't think we
want anything that tenuous.  In general, not all services are
client-server.  Second, not all responses are to be viewed: some are
machine to machine (so nobody views them) and some are non-visual (SIP
calls, for instance).  Third, not all Internet services are
connnection-oriented: UDP datagrams, for instance, are connectionless.
Finally, for a given service and a given inbound datagram, response
datagrams might or might not be generated depending on various local
policies.  It's problems like this that make people avoid trying to
define too precisely.

It does seem that a service on the Internet is something that accepts
datagrams, when those datagrams are not necessarily the result of
datagrams sent by the same thing.  Therefore, I _think_ this will
work:

    services (i.e., any software process that accepts datagrams from
    the Internet, when those datagrams are not themselves necessarily
    the consequence of a datagram previously sent by the software
    process itself) that use the Internet’s unique identifiers

I'm not absolutely sure this is right, but I think it might be close
enough.  The problem with it, of course, is that every single thing
connected to the Internet uses at least one of the Internet's unique
identifiers (this problem is, remember, part of what alarmed the IAB
in the first place).  So this basically says, "Any software process
that accepts datagrams from the Internet," and that very nearly boils
down to, "A program with a socket open to the Internet."  If that's
ok, I guess I can live with it, though it'd be good to get some more
technically-clueful (or even, some would argue, "at least one pair
of") eyes on this.

Best regards,

A

-- 
Andrew Sullivan
ajs at anvilwalrusden.com