Ted Lemon wrote:
On Nov 4, 2013, at 1:01 PM, Dave Cridland <dave(_at_)cridland(_dot_)net>
wrote:
What Larry is saying - or at least, what I'm agreeing with,
regardless of whether Larry meant this or not - is that if
you've a bunch of people saying "I've implemented this in
production, and X needs to be Y", then that is a very hard
argument to beat.
Historically we haven't just blessed existing implementations
if there were things that needed to be fixed. This interpretation
of "running code" is a bit problematic because people with running
code have a vested interest in the spec not changing.
Not always a bad thing, but it can be quite harmful.
It has proven to be quite harmful several times and in several ways.
It regularly causes bloat (functionality that is easily dispensible
to a huge amount of users/consumers) to stay in the specification with
a requirements tagged to it, rather than being a real option.
And the folks fighting for the bloat to stay in the spec are either
vendors that appreciate when its harder for competition to compete
in general, rather than for a limited subset of consumers. The others
are organizational interest groups (often governmental) that, while
representing a minority among real consumers, would like specific bloat
in products without having to pay extra. I don't mind a standard
offering them _options_. But I strongly dislike the IETF pushing
requirements on bloat that the large majority doesn't need,
just to please a specific minority.
-Martin