On Mon, Nov 4, 2013 at 11:28 AM, Ted Lemon
<Ted(_dot_)Lemon(_at_)nominum(_dot_)com> wrote:
On Nov 4, 2013, at 8:07 AM, Peter Saint-Andre <stpeter(_at_)stpeter(_dot_)im>
wrote:
For me, "rough consensus" and "running code" should be taken
together, not independently. I've always taken it as "rough consensus
OF THOSE WITH running code".
Larry, that is awesome!
Unfortunately it's pretty easy to game this by writing some running code
that isn't very usable, and then claim that you are right based on the fact
that you have running code and the other folks don't. I think this should
certainly be considered, but it's not quite as strong an argument as is
being claimed.
Because of this, I would weight an open source implementation more heavily
because it can be viewed by the community. I've found that developing an
open source implementation at the same time as the draft is under
development has been highly valuable to both; the running code can be
inspected for quality, and what's left after all of that gets tightened
thus deserves reflection in the draft revisions.
I realize it's easier to talk about open source in applications work versus
the lower layers, however.
Also valuable are interoperability events; even if the source code can't be
shared, bogus implementations become readily visible.
I think the output of both of those projects deserve serious consideration
when evaluating consensus. If rough consensus alone is clearly in
contradiction to the data provided by running code, there's something
seriously wrong.
-MSK