Re: The IETF process

Marc Hedlund writes:
	...

 > I am sympathetic to Shel's concerns and (since I say such foolish things
							     ^^^^^^^^^^^^^^
Hey!

	...
 > It would be foolhardy for an author to submit, or a standards body to
 > accept as final, a specification that has never been implemented.  The very
 > best, simplest ideas can go straight to hell when you sit down to convince
 > a computer of their elegence, and that learning process will necessarily
 > improve the final specification.
 > 

Don't get me wrong. I could not agree with you more.  As I wrote in my
message, I understand the need for implementation before
standardization.  The problem I pointed out is that some people who
have good ideas don't have ready access to test beds, or are simply
more focussed on other things, and http-wg has been the closest
approximation many of us have to a place to discuss the ideas.  Now
that the NCSA folks have offered to consider implementing ideas
suggested to them, that may help resolve the catch-22 I described.

 > That does _not_ mean, to my mind, that an idea must be formulated,
 > specified, and tested before it is discussed on this list.

But it does mean that after the discussion, nothing happens without an
implementation, and there has often been no clear link from the
group's discussions (which often terminate without a consensus or
decision anyway) to follow-on implementations.  There is also the minor
matter that it is work to do implementations, and nobody can be
expected to volunteer to do that work.  Yet if people are expected to
implement everything before bringing it to the WG, that again gets
back to the same problem.  If research organizations such as NCSA want
to foster this activity that seems to be a good solution.


  I would
 > consider it perfectly reasonable to see a note which read, "This is my
 > idea, I think it would improve the spec, what do you think?"

Such discussions frequently seem to end without consensus or decision.
That is something the WG *can* do something about.

 If response is
 > favorable, a follow-up could be, "Can someone throw together a libwww-perl
 > (or whatever) implementation to see if it would work?"

A problem I have experienced repeatedly in http-wg is that many
excellent proposals seem to be discussed only partially, then dropped.
And I am not just talking about my own proposals either.  For example,
Jeff Mogul's "GET if-validator-valid" proposal was never clearly
adopted or rejected, much less put to the test (as far as I know).
Without clear closure followed by a request that someone volunteer to
do a test implementation there is a real limit to progress.  It's
inevitable that not all participants on this list care about all
issues, but there ought to be some way to decide (or at least
*detect*) if a proposal has enough support to warrant an
implementation.  Also, some proposals do not require implementation,
as they are merely a matter of clarifying the spec in areas where
different implementations behave differently.  These also often seem
to go nowhere.

 > 
 > One other quote from Roy:
 > 
 > >   1) People on this list (and the other Application-area WGs)
 > >      tend to "blab" a little too much and "make concrete proposals"
 > >      too infrequently.
 > 
 > These comments encourage wg members to get down to brass tacks more
 > quickly.  We've had an endless string of endless strings of messages, and I
 > was startled to see an actual polling of list members resulting in a clear
 > picture of list opinion (Host v. Orig-URI) last week.

Yes, it would be great if we did this more often.

	...

 > 
 > Regards,
 > Marc Hedlund <hedlund@best.com>
 > 

Shel Kaphan

Received on Tuesday, 26 September 1995 12:14:02 UTC