W3C home > Mailing lists > Public > w3c-dist-auth@w3.org > April to June 2001

RE: "nothing left to cut" (was: Re: [ACL] Owner issues)

From: Clemm, Geoff <gclemm@rational.com>
Date: Wed, 6 Jun 2001 11:06:52 -0400
Message-ID: <3906C56A7BD1F54593344C05BD1374B1033E59E6@SUS-MA1IT01>
To: acl@webdav.org, ietf-dav-versioning@w3.org, w3c-dist-auth@w3.org
There are several very different issues here.

The "cut until there is nothing left to cut" refers to protocol semantics
(i.e. what a server is required to do), *not* the text in the protocol

It is of course essential to supplement the normative statements
(i.e. those with a "MUST") with as much descriptive text as is
necessary to make the meaning clear.  In particular, with descriptive
text, it is better to err on the side of "too much" than "too little".
So the criteria for descriptive text is "keep adding as long as the
net clarity of the document is increased".  (But there are
additions that can decrease the net clarity of the document, so
the criteria is not just "add as much as possible").

I believe that "minimizing the protocol semantics" and "maximizing
the protocol document descriptive text" are in fact complementary goals.

There is another issue, of whether you repeat the same normative
statement in different ways in different parts of the document.
In this case, I do believe that a particular normative statement should
in exactly one place, but that descriptive text about that normative
statement can and should appear in several places in the document
(i.e. wherever it is useful to do so).


-----Original Message-----
From: Greg Stein [mailto:gstein@lyra.org]
Sent: Wednesday, June 06, 2001 2:07 AM
To: Clemm, Geoff
Cc: acl@webdav.org; ietf-dav-versioning@w3.org; w3c-dist-auth@w3.org
Subject: "nothing left to cut" (was: Re: [ACL] Owner issues)

On Tue, Jun 05, 2001 at 01:26:43AM -0400, Clemm, Geoff wrote:
> I believe I was the one that originally suggested we allow
> updating the owner with a PROPPATCH.  I have seen the error
> of my ways (:-).  So unless there really is someone that
> feels this functionality is important, I believe that the
> principle of "you aren't
> done until there is nothing left to cut" says to take it out.

I am beginning to seriously disagree with the whole notion of "cut
everything until there is nothing left to cut."

You are taking it to the extreme, leaving a specification that is obtuse,
hard to understand, and requires a half-dozen readings just to figure out
the subtleties and interactions between the elements, such that you can
*infer* what should have been outright specified.

Cutting features is great. Creating obtuse specifications is absurd.

If you want a *STANDARD*, then it must be obvious to *all* implementors what
the standard should be. If one out of twenty people can figure out ALL of
the implications and inferences to implement the "standard", then you simply
DON'T have a standard. You've only created a guide. The other 19 people
implemented something wrong because they couldn't grok the darned document.

I'm not making a statement on the <owner> thing. Instead, I'm arguing that
your policy is erroneous. It needs to be tempered.

[ I believe this applies more to the DeltaV spec than the ACL spec (I
  haven't read the ACL spec lately); the DeltaV spec is currently a very
  opaque document because of the "say it once; anything more is redundant"
  attitude taken towards it. ]


Greg Stein, http://www.lyra.org/
Received on Wednesday, 6 June 2001 11:02:54 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 20:01:22 UTC