Re: Evolving Documents (nee "Living Documents") side meeting at IETF105.)

[Phillip Hallam-Baker <[e-mail]>]

On Thu, Jul 18, 2019 at 10:02 PM Ted Lemon <[e-mail]> wrote:

> On Jul 18, 2019, at 9:50 PM, Keith Moore <[e-mail]>
> wrote:
>
> Yes, and I’ve repeatedly said I could see optimizing in corner cases..
> But I think it’s a rare WG that doesn’t have any potential to adversely
> affect other interests.
>
>
> Another way to look at this is a well-known cognitive bias: “I am right.”
>   If you look at what a working group is doing and don’t understand it,
> there is a tendency to think they don’t know what they are doing, and that
> you know what they should have done.   This bias is frequently wrong, and
> I’ve seen it turned against good work numerous times.
>

Thing is that sometimes the criticism is correct but irrelevant. If you are
fixing a 30 year old protocol to respond to a need that wasn't originally
anticipated, there are going to be times when you have to break things. And
very often the pushback that comes is of the form 'there are still people
using fidonet who will be inconvenienced', to which my response is
generally of the form, 'this is not a historical re-enactiment society, we
do not support retro-computing.'

There is a bias in favor of supporting the oldest established communities
rather than the needs of the largest communities of users. The Internet now
has 4 billion people using it and there is a tendency for some folk to
prioritize the perceived needs of legacy users that may not even exist.

The other pathological behavior I see very often is people insisting that
the real problem is deployment who also insist on technical approaches that
make deployment impossible.

I spend more time thinking about deployment of my protocol designs than the
protocol itself. I look at the business models of each of the parties that
I need for adoption. For the Web, I was running simulations of different
deployment scenarios.

Which is why I find statements of the form 'we have to move fast on this so
we will require the deployment of a new TCP feature requiring kernel level
support' farcical.