On Apr 20, 2012, at 5:40 AM, Andrew Sullivan wrote:
On Fri, Apr 20, 2012 at 12:17:46AM +0300, Yoav Nir wrote:
The "Experimental" designation typically denotes a specification that
is part of some research or development effort.
However, I do not believe that this is still typical. Authors come up with
ideas that they think are useful. If when the documents are ready for
publication it is still not clear whether there are enough implementers
convinced by the use case (some definition of running code), they are
encouraged to publish them as experimental rather than proposed standard.
Yes, and surely what we want is that we continue to mislabel as many
RFCs as possible in this way.
No, I don't believe they are mislabeled, they just don't fit the statement in
RFC 2026, and they're not scientific experiments. It gives fair warning to
would-be implementers regarding the maturity of the specification. Other
implementations may not have implemented this RFC (if it's an extension), or
they may solve the same problem with some other mechanism. Or they may even
have implemented the same RFC, but we're not sure yet that it's unambiguous
enough and error-free enough to produce interoperable implementations.
We want to make our own contribution to
the gradual reduction of all words to mean everything else, so that we
can all communicate exclusively by mumbling, "Whatever." Now that the
AP Stylebook has given up on "hopefully" and Oxford is playing with
the comma, anything goes.
Neither of these can really be called an "experiment" except in the
sense of "let's see if the Internet needs this specification".
Yes, and that's a stupid meaning of the term in this case, because
quite frankly all the rest of the RFCs are also experiments in this
sense (and many of them have delivered the clarion answer, "Don't need
it at all!" When was the last time you used BEEP?)
I don't know what I think about the proposal overall, but I will say
that one thing definitely in its favour is that it would discourage
those who think that "Experimental" is the Standards Track Consolation
Prize Track. I'd be wholeheartedly in its favour, except that the old
"2 year review" standards track process didn't work, so we ditched
it. I don't see how recycling the approach with Experimental
documents is any more likely to succeed.
True. Perhaps if successful standards were advanced form Proposed to something
else, there would be less pushback against publishing ideas with little running
code as PS. Surely your little idea with no running code can't be in the same
category as TLS, right?