[ih] Why did TCP win? [Re: Internet-history Digest, Vol 63, Issue 3

[Dave Crocker]

> I think your comment about "unanimity" is a bit misleading. As I 
> recall, there were always some people who wanted more, but the 
> leadership assumed its burden and said No, and everyone respected them 
> enough to leave it at that. 


The 'rough consensus' model dominated Arpanet and Internet work, where 
that 'strongly dominant agreement' made the decisions, quite a bit more 
than an authority's breaking a logjam.  (Again, I said 'more', rather 
than always or never.)

There are always people who want more.  There are always people who are 
never satisfied.  That why 'rough' is such an apt word, and especially 
given that it has two meanings, both of which apply.

I frequently cite Bob Hinden's example of a working group's being 
deadlocked.  At a meeting, he did yet-another round of hand-raisings for 
either of the two preferred choices and it was, yet again, clear that 
neither one was strongly dominant.

He said he asked who wanted resolution of this, and nearly all the hand 
went up.

He said he then noted the lack of wg progress and offered that 
resolution would not be possible unless enough folk changed their choice.

He then did another round of hand-raising and one of the choices was 
finally dominant.

I did a variant of this years later, at MAAWG, for the DKIM revision.  
DKIM has the task of delivering an authenticated domain name, to 
indicate its owner takes 'some' responsibility for a message. This 
defines a relatively noise-free message stream, for reputation analysis.

But we made an error in the original specification and provided two 
places for specifying a domain name, with no indication which was the 
one to be used for assessments.

The community debated the choice for a very long time, with no 
resolution.  I reviewed the online discussion and solicited a small 
group from it to have a separate exchange.  I selected people based on 
their willingness to discuss things reasonably, rather than because they 
agreed with my own preference.  The resulting set of folk actually 
tended toward the choice I didn't want, but at this point, some 
resolution was better than none.

My role was strictly as facilitator.  (I started by noting my preference 
only once, just to acknowledge it.) The group had a healthy debate and 
did reach a rough consensus.  That group then went back to the main 
discussion and the force of a coherent preference among a core of strong 
contributors got the rest of the community to resolve the choice.

Specifications succeed because they work and they are supported and they 
get built and they get deployed.

No formal authority can guarantee that that entire sequence will happen, 
because there is no authority with that much power.

Except the authority of the community.

d/

-- 
Dave Crocker

Brandenburg InternetWorking
bbiw.net
bluesky: @dcrocker.bsky.social
mast: @dcrocker at mastodon.social