Properties tests based on invalid assumptions?

[Joe Orton]

On Fri, Oct 31, 2008 at 06:37:13PM -0700, John Meissen wrote:
...
> RFC4918, section 9.2 says
> 
>  All DAV-compliant resources MUST support the PROPPATCH method and MUST 
>  process instructions that are specified using the propertyupdate, set, 
>  and remove XML elements. Execution of the directives in this method is, 
>  of course, subject to access control constraints. DAV-compliant resources
>  SHOULD support the setting of arbitrary dead properties.
> 
> The key words here are that the resource MUST support the method, and
> MUST process the instructions; the resource SHOULD support the setting
> of arbitrary properties.
>
> In my interpretation that means that I must implement the method and
> parse/process the body of the request, but I don't have to actually
> support anything other than the live properties associated with the
> resource. And changes to live properties are subject to constraints.
> 
> So failing to set arbitrary properties is, in fact, a successful
> response, so long as the response is constructed correctly.

I missed this message, sorry.

I think that the correct behaviour for a server which doesn't support 
arbitrary properties would be to return a 207 for the PROPPATCH which 
gives a propstat with status 403 for each "set" instruction which is not 
supported.

The way the litmus code is currently written it should FAIL for such a 
response; subsequent tests should then SKIP.  Is that what you're 
seeing? 

It's reasonable to argue that the propset FAIL should be a SKIP since 
it's DAV-compliant to not support arbitrary dead props, though I might 
argue that it would be simpler for people testing such servers to simply 
skip running src/props at all.  It would also add quite a bit of 
complexity to litmus to handle this gracefully.

If you are saying that your server pretends that a propset succeeds when 
it is actually being silently ignored, I don't think that would be DAV 
compliant at all.

Regards, Joe