Re: iCAP, OPES and the IETF - resend

Roy,

First, a few factual corrections.

* If you are / were a member of the iCAP forum then you received the first 
draft in Feb 2000; some *2* months after the release, not 8 as you state. 
(Though perhaps you are referring to the later publication of the 
Internet-Draft specifically for the Pittsburgh IETF?)
* The group "pwg" has existed since the iCAP forum was formed and was not 
formed after the release of the I-D.
* Discussion was never "moved to WREC" - iCAP was never part of the WREC 
charter.
* For the record, here are the facts I can determine from my mail archives.

- Early Dec 1999 - press release from Akamai, NetApp and others
- Nov-Jan 2000   - editors group expanded - pwg formed
- Early Feb 2000 - first public draft to those registering on the web site
- Early Mar 2000 - decision by tech editors not to submit BoF request at IETF;
                   we were too late and there was some concerns about the
                   benefits of having something at short notice in Australia
                   (since many of those would have gone for WREC which did not
                    meet in Adelaide).

***- Early Mar 2000 - first iCAP I-D? (I can't find the exact date)

- Early May 2000 - BoF at N+I, (Vegas?)
- Mid May 2000   - BoF at Web Caching Workshop in Lisbon
- Late June 2000 - shaky but running code from netapp and a few others
- August 2000    - iCAP on agenda at WREC WG (rather than separate BoF). Note
                   that iCAP was never part of WREC's charter. iCAP also 
discussed
                   at Content-Adaption BoF.

...etc.

For the rest, yes, iCAP was originally designed to be HTTP encapsulated 
over HTTP but implementation showed this to be impractical and unperformant 
- particularly for large objects where only partial objects needed 
vectoring (e.g. in virus scanning). So, after a lot of discussion, the 
decision was made to remove the requirement for HTTP compatibility but 
retain the header-body syntax a la HTTP (and MIME).
Yes, you are unlikely to see an implementation of the current spec because, 
as I've said before, it is a superset of all the changes made by all those 
who've implemented iCAP v0.9 and fixed bugs and inconsistencies along the 
way. So, it is indeed a documentation of "iCAP-as-implemented" in that it 
is a record of what people have coded.
That aside, we are planning to release reference code for both the client 
and the server for iCAP v1.0.
Finally, I agree 100% with your last paragraph (excluding the last 
sentance, of course) - continuing the publication process is exactly our 
intention. Where this discussion takes place is still, it seems, moot. 
There are those that want it within OPES and those that don't. Personally, 
I don't care. I have proposed that we discuss iCAP development on an iCAP 
list for the time being, at least until we have a home elsewhere - 
preferably within an ietf wg.
John

At 04:05 PM 11/06/01 -0700, Roy T. Fielding wrote:
> What iCAP protocol?  You may not be able to see this progression from the
> inside of NetApp, but here is the status of iCAP from the viewpoint of an
> independent implementer of the specification and a member of the iCAP forum:
>
>    1) A press release was made
>
>    2) Roughly eight months later, a specification was sent out as a draft
>
>    3) A bunch of organizations tried to implement according to that draft
>       and found that it was hopelessly screwed for performance
>
>    4) Several of the implementers came up with independent theories on
>       how best to correct those issues, some variations of which were
>       packaged within beta releases of system software
>
>    5) The pwg was formed to make some coherence out of those changes into
>       a protocol draft to be submitted as an Internet Draft (v0.9)
>
>    6) More experience implementing the protocol uncovered differing opinions
>       as to what the spec meant and why it was no longer HTTP compliant
>       even though it was (at the time) attempting to be.
>
>    7) Discussion moved to WREC, where the protocol was completely changed
>       to disregard HTTP compatibility but continue to use the least desirable
>       aspects of HTTP syntax, several spec errors were fixed (and more
>       introduced), and a draft was issued as iCAP 1.0.
>
>    8) Discussion continued regarding both minor spec errors and major
>       design issues, extending from WREC to OPES.
>
> In other words, I have no idea which iCAP you are claiming to be ready for
> publication as an RFC.  I have yet to see a single implementation of iCAP 1.0
> as defined by the current specification.  Judging from the comments discussed,
> that draft contains many inconsistencies and incompatibilities with even
> NetApp's implementation of the protocol.  So, what on earth gives you the
> right to declare that draft as a representation of iCAP-as-implemented?
>
> Continue with the current publication process.  Edit the specification to
> reflect what has been implemented and identify what versions of what software
> are considered implementations of the protocol, so that we can independently
> verify that the specification is correct.  Then we can decide whether or not
> it is ready for publication as an RFC, and with what status the RFC should
> be labeled.  Feel free to ignore the OPES functionality and requirements for
> this version of iCAP, but publishing a specification that matches nobody's
> implementation exceeds my threshold of IETF abuse.
>
> ....Roy
---------------------------------------------------------------
Network Appliance           Direct / Voicemail: +31 23 567 9615
Scorpius 2                                 Fax: +31 23 567 9699
NL-2132 LR Hoofddorp               Main Office: +31 23 567 9600
---------------------------------------------------------------