Re: Issue 147 - Depth vs Conditional headers

I don't see an alternative.  The only thing I'd suggest is to
slightly simplify the qualifier by not referring to the If header
(since we can state in the definition of If header that it
doesn't recurse via Depth), i.e.:

"Any headers on a method that has a defined interaction with the Depth 
header MUST be applied to all resources in the scope of the method 
except where alternative behavior is explicitly defined, and except for 
the If-* headers defined in RFC2616."

Cheers,
Geoff

Julian wrote on 01/07/2006 07:26:11 AM:
> 
> See <http://ietf.cse.ucsc.edu:8080/bugzilla/show_bug.cgi?id=147>.
> 
> Hi,
> 
> it seems that RFC2518 and RFC2518bis share a major problem in defining 
> how the Depth header and other headers interact. In Section 9.2, RFC2518 

> (<http://greenbytes.de/tech/webdav/rfc2518.html#rfc.section.9.2.p.9>) 
says:
> 
> "Any headers on a method that has a defined interaction with the Depth 
> header MUST be applied to all resources in the scope of the method 
> except where alternative behavior is explicitly defined. For example, an 

> If-Match header will have its value applied against every resource in 
> the method's scope and will cause the method to fail if the header fails 

> to match."
> 
> Let's look at an example. A COPY request with Depth infinity having an 
> If-Match header such as:
> 
>    COPY /foo HTTP/1.1
>    Host: example.com
>    Destination: http://example.com/bar
>    If-Match: "123"
>    Depth: Infinity
> 
> would need to evaluate the If-Match header for each resource to be 
> copied. There are two major problems with this:
> 
> 1) It would be extremely expensive. For a copy of a large tree, the 
> server would need check entity tags on any member before even starting 
> the copy operation, because it would need to fail with a top-level 412 
> status otherwise.
> 
> 2) It's not being implemented in today's servers, see 
> <http://ietf.cse.ucsc.edu:8080/bugzilla/show_bug.cgi?id=147#c3>.
> 
> 
> Now for the "If" header itself (this is the only applicable header 
> defined in RFC2518 itself), RFC2518bis already states that it only 
> applies to the Request-URI (when the untagged form is used).
> 
> 
> The only simple fix I can currently think of for the Depth header 
> definition though is to revert the definition at least for those headers 

> where we know that servers aren't doing it as defined (and also are not 
> likely to change their behavior). Such as replace the quoted paragraph 
by:
> 
> "Any headers on a method that has a defined interaction with the Depth 
> header MUST be applied to all resources in the scope of the method 
> except where alternative behavior is explicitly defined, and except for 
> the If-Dash headers defined in RFC2616 and the If header defined in 
> Section x.xy)."
> 
> Feedback appreciated (I mean it! :-),
> 
> Julian
> 
> 
> 
> 
> 
> 
>