Re: [Gen-art] Gen-ART review of draft-ietf-vcarddav-webdav-mkcol-05

[Cyrus Daboo <[e-mail]>]

Hi Spencer,
Thanks for your review - I was on vacation last week so I am only getting 
around to replying now.

--On August 7, 2009 6:36:41 AM -0500 Spencer Dawkins 
<[e-mail]> wrote:

> Comments identified as "Spencer (clarity)" are provided for editors, but
> are not part of the Gen-ART review.
>
> Abstract
>
>    This specification extends the Web Distributed Authoring and
>    Versioning (WebDAV) MKCOL method to allow collections of arbitrary
>    resourcetype to be created and to allow properties to be set at the
>    same time.
>
> Spencer (clarity): I know that RFC 4918 didn't provide an expansion for
> MKCOL until about page 46 (or maybe 25, in passing), but since MKCOL
> appears in the abstract for this document, it might be helpful to
> s/MKCOL/MKCOL ("Make Collection")/ here. MKCOL is defined at first use in
> the Introduction, so that's fine - this is only a suggestion about the
> abstract.

Sure - seems reasonable. Change done in my working copy.

> 3.  WebDAV extended MKCOL
>
>    The WebDAV MKCOL request is extended to allow the inclusion of a
>    request body.  The request body is an XML document containing a
>    single DAV:mkcol XML element as the root element.  The Content-Type
>
> Spencer (minor): if I'm reading this paragraph correctly, I'd suggest
> "The request body is an XML document that MUST contain a single DAV:mkcol
> XML element as the root element" here - the last sentence in this
> paragraph makes me think the requirement is normative, but it doesn't
> look normative to 2119 scanners :-)

As per comments from Julian I won't make this change.

>    request header MUST be set appropriately for an XML body (e.g., set
>    to "text/xml" or "application/xml").  XML-typed bodies for an MKCOL
>    request that do not have DAV:mkcol as the root element are reserved
>    for future usage.
>
>    As per the PROPPATCH method ([RFC4918], Section 9.2), servers MUST
>    process any DAV:set instructions in document order (an exception to
>    the normal rule that ordering is irrelevant).  Instructions MUST
>    either all be executed or none executed.  Thus, if any error occurs
>
> Spencer (clarity): this sentence reads roughly to me, and it's 2119
> language, so needs to be clear. I suggest "The set of instructions MUST
> be executed atomically; either all are executed or none are executed".

The text here is an exact copy of that in 4918. Strictly speaking all 
instructions are executed, it is just that some succeed and some fail. 
Perhaps better text is:

"If any one instruction fails to execute successfully, then all 
instructions MUST fail (i.e., either all succeed or all fail)".

>    during processing, all executed instructions MUST be undone and a
>    proper error result returned.  Failure to set a property value on the
>    collection MUST result in a failure of the overall MKCOL request -
>    i.e. the collection is not created.
>
> 3.5.  Example: Successful Extended MKCOL Request
>
> Spencer (minor): Perhaps this should be s/Successful/Unsuccessful/? It's
> returning a 403/424 :D

Fixed.

> 4.  Replacing Existing MKxxx Methods
>
> Spencer (clarity): Hmm. This section (and 4.1, and 4.1.1) aren't quite
> about REPLACING existing MKxxx methods, but more like "Providing MKxxx
> functionality using extended MKCOL". Would that be clearer? The current
> text in 4.1 sent me looking to see whether "replacement" meant
> "deprecation" (it doesn't, if I'm reading clearly), for example.

Well ultimately I would like to see MKCALENDAR deprecated in favor of 
extended MKCOL - however, given the ever growing deployments of CalDAV that 
is probably not going to happen any time soon.

So, I have changed the title to:

"Using Extended MKCOL as an Alternative to MKxxx Methods"

and changed "replace" to "alternative for" in a couple of places.


-- 
Cyrus Daboo