Document Series

[Pete Resnick]

On 6/4/03 at 2:44 PM -0400, Keith Moore wrote:

>what I mean is that if we realize that PS in effect is taken as 
>"ready for deployment" (which seems to be how the industry 
>interprets it), then maybe we really do need to expect some level of 
>implementation and interop testing before we approve something as PS.

So, I read this as saying that industry is interpreting PS as what we 
meant DS to mean. That is, industry thinks:

   A [Proposed] Standard [is] well-understood and known to be quite
   stable, both in its semantics and as a basis for developing an
   implementation. A [Proposed] Standard may still require additional or
   more widespread field experience, since it is possible for
   implementations based on [Proposed] Standard specifications to
   demonstrate unforeseen behavior when subjected to large-scale use in
   production environments.

   A [Proposed] Standard is normally considered to be a final
   specification, and changes are likely to be made only to solve
   specific problems encountered. In most circumstances, it is
   reasonable for vendors to deploy implementations of [Proposed]
   Standards into a disruption sensitive environment.

[2026, section 4.1.2 on Draft Standard, with "industry" edits.]

I don't exactly understand why this should mean that we should (as we 
seem to be on our way to doing unofficially already) raise the bar of 
PS to be identical to the bar of DS unless we really think that there 
is no worth in having Proposed Standard as it was originally intended 
(a vetted, though immature, specification). We could go the route 
(similar to what CMH suggested) of going Experimental->PS->S, or even 
Experimental->DS->S, but I assume that the only thing that this will 
result in is industry thinking that Experimental is ready for prime 
time. (They already assume that "Informational" is an "IETF-Approved 
implementable standard".)

Wouldn't it make more sense to:

- Leave our written processes exactly the way they are;
- *Lower* the bar on PS back to what it was intended to be, get PS's 
out early, and do real interop testing early in the process;
- Start putting a boilerplate on PS RFC's which says "OK, now we 
really mean it; this isn't ready for prime time!"
- Ignore all screams of horror when we significantly change a 
specification that is at PS, publish the changed spec as a new PS and 
change the old PS to Historic, and say "I told you so".

I'm not being facetious. Or at least, I'm not being totally 
facetious. If the problem we are addressing is really, "Industry 
expects more out of PS than we intended", the problem can be 
addressed by either changing our interpretation of PS to meet 
industry expectations, or by changing industry expectations. The 
former seems to be what the IESG has been doing for some time now, 
and the result has been that some industry folks have taken to 
considering Internet Drafts stable enough to implement. This seems 
destined for an arms race. I'm more inclined to attempt the latter if 
we can.

pr
-- 
Pete Resnick <mailto:presnick at qualcomm.com>
QUALCOMM Incorporated - Direct phone: (858)651-4478, Fax: (858)651-1102