Hi -
From: Miles Fidelman <mfidelman(_at_)meetinghouse(_dot_)net>
Sent: May 26, 2014 7:18 PM
To: ietf(_at_)ietf(_dot_)org
Subject: Re: inquiry re. the state of protocol R&D
I have to vehemently disagree. To me, APIs are a step in the wrong
direction.
Protocol specs - framed as PDU formats and state machine models -
present a basis for interoperability and distributed operation. APIs are
language-specific, and all too often are tied to a
centralized/client-server model of the world. A big step backwards.
It depends on what folks mean by "APIs". My experience in the IETF
is that when the term is used here it frequently does *not* mean
"language binding", even though that seems to be the term's usual
sense outside the IETF. The sense here frequently seems to be
something more akin to "service definition" and not tied to a
specific language binding, but I haven't seen it rigorously
spelled out here.
The SNMPv3 work did define internal "ASIs" (abstract service
interfaces) helpful in characterizing necessary computations
behind protocol behavior, but these don't describe the services
provided by the protocol in a way that would be particularly
helpful to its users, whether human or software.
Randy