Re: Issue 455 closed: Representation header and SOAP processing model

Noah, 

it seems to me that what you are describing is the default behavior -
Representation header is removed by any node that processes it, except
when the node knows better, e.g. by following the rules of our sticky
role.

So, except when using the sticky node, other features or an active
intermediary, Representation header gets dropped when processed.

Basically, the default expectation is that a role is played by a single
node. The soap:relay attribute widens this to the expectation that a
header targeted at a role is consumed at a single node playing that role
(the first one that chooses to process the header). Then it's up to the
headers or other features to change the expectation. With our sticky
role we've done enough for Representation, I believe. 8-)

Jacek


On Fri, 2004-03-19 at 16:59, noah_mendelsohn@us.ibm.com wrote:
> Jackek Kopecky writes:
> 
> >> I wouldn't like the text to suggest that 
> >> inventing roles for various relaying 
> >> semantics is necessarily a good practice 
> >> because I have a feeling these would 
> >> mostly be one-off deployment-specific specs.
> 
> I agree.  The case I had in mind is that people will invent roles that 
> basically mean "the node named A".  In other words, the common case where 
> you know that a role is designed to be played by exactly one node.  I am 
> not proposing to generalize that in SOAP, but I'm wondering if it doesn't 
> make sense in the recommendations for particular headers to say: "if you 
> recognize the role (as you probably do if you're assuming the role), and 
> you have knowledge that the role is designed to be played by exactly one 
> node, then you MAY/MUST/SHOULD decline to relay the header. "  We already 
> say that the spec for a header can set relay rules for a header that is 
> processed, so I think that such a specification is at least allowed by the 
> SOAP processing model.  I can see both sides of the question as to whether 
> we should do it for the representation header in particular.
> 
> One question:  if we don't allow this, then why invent a "sticky" role? 
> They would all be sticky, no?
> 
> --------------------------------------
> Noah Mendelsohn 
> IBM Corporation
> One Rogers Street
> Cambridge, MA 02142
> 1-617-693-4036
> --------------------------------------
> 
> 
> 
> 

Received on Friday, 19 March 2004 11:12:39 UTC