[Edm] Newcomer to the list

[Edward Lewis <[e-mail]>]

On IETF 118's agenda I saw "Evolvability, Deployability, & Maintainability" on the agenda and thought this sounded like it could be interesting.  I've been a long-time attendee of the IETF (1994-2016 and again in 2023), the 6-year gap has left me a little rusty when it comes to the meeting websites, but I was able to find the agenda for the meeting (which didn't yield a lot of info on what would be in the session).

I couldn't find (and still haven't found) a charter, just a group description.  (Maybe that is the charter.)  The meeting had no introductory material, so I wasn't able to get oriented in the room.  Since then, I've tracked down the one document that was being discussed - "Maintaining Protocols Using Grease and Variability" and gave it a quick read.  I'm mentioning this to provide an excuse for being naïve about the discussion.

The reason I was interested in the title is that I've been measuring (DNS top-level domain) operator deployment of DNSSEC for over 10 years and RPKI for just about 4 years.  (I've been involved with DNSSEC deployment for 25 total years, my data is only since 2011.)  One of the early goals of my measurements was initially to determine where to promote the technologies in the DNS economy.  But this summer it struck me that the operators weren't in need of more education, training, or encouragement, the low deployment numbers were a sign that operators did not accept these protocols as ready for operational use.

My comment on the meeting (I attended just the 1st-half, I was called away) and the draft is that I have not heard nor read any consideration of network operators.  To me, deployment is when an operator puts a protocol into service, not when there are available implementations.  For example, in DNS - we have four well-known, widely deployed opensource implementations of the DNS protocol plus a handful of well-recognized proprietary implementations, and countless other ad-hoc implementations.  We have lots of code bases.  But many DNS features are simply not used, DNSSEC signing of zones is hovering in the single digits, percent wise, even after 25 years of deployment effort.

I've thought, informally, about proposing that operational deployment metrics are a means to measure whether a protocol (enhancement) is a success, not only relying on code availability.  This would encourage consideration, in protocol design, of what would make a protocol truly address the problem it set out to solve.  The IETF has a long tradition of protocol engineering with an eye on interoperable implementations as the end result, I don't know if stretching to consider whether the operational quality of a protocol is within scope or not.

As an aside - Operational quality includes aspects if whether the protocol is manageable - meaning it can be clearly understood, can be measured, can be debugged, can be updated (evolved) in place, etc.