[rfc-i] I need your "good" RFCs

Joseph Touch touch at strayalpha.com
Tue Feb 16 14:36:25 PST 2021



> On Feb 16, 2021, at 1:31 AM, tom petch <daedulus at btconnect.com> wrote:
> 
> I said earlier that TCP and the trio of SNMPv1 RFC were high on my list.  I think it worthwhile to say what are not, and top of that list would be the original, core specifications of IPv6. 

There are many on my list to avoid - starting, first and foremost, those that acceded to the directive of “no API in a protocol spec”.

A protocol spec without an (abstract) API isn’t a protocol spec.

In my courseware (for intro to nets), I define a protocol spec as a finite state machine (FSM) as follows:

	- list of states
	- set of upper layer inputs/outputs (the abstract API)
	- set of messages transmitted/received (the “on the wire” format)
	- set of timers to be set or trigger 
	- a table that maps every input (API input, message receipt, and/or timer trigger) and state combination to a new state and set of outputs (API output, message sent, and/or timer set)

PS - “protocol spec as FSM" isn’t merely a convenience or convention. It’s fundamental, based on what you get when you take a Turing machine and connect it to a communication channel, i.e., a FSM is a Turing machine that can’t “un-send” its outputs.

Joe

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.rfc-editor.org/pipermail/rfc-interest/attachments/20210216/3e24a8c3/attachment.html>


More information about the rfc-interest mailing list