Re: Why is [message id] required for requests but not for other messages?

We are currently trying to define the semantics of [message id], and I
would like to understand the use cases.

Right now we have one real use case, namely message correlation.

We also have a bunch of ideas that might generate use cases.  We might
want a message id for logging.  We might want it for duplicate
elimination.  We might want it for detecting replay attacks, and so
forth.  We've done a bit of exploring as to what requirements might come
out of these possible use cases.  We seem to be backing away from
supporting any particular one of those, and this seems wise in the light
of last week's email traffic.

Right now the rules under consideration appear to be:

   1. [message id] must be present in requests (status quo, re-affirmed
      by closing of LC86)
   2. A reply must relate-to the [message id] of the request that
      generated it. (status quo)
   3. Anything that gets a message (request or otherwise) with a
      [message id] it's seen before may do anything at all with that
      message. (under consideration)

That rule 3 is actually a pretty big hammer.  To be safe, I would
personally use a GUID wherever a [message id] was mandated and leave it
out entirely otherwise.

What use case are these rules supporting?  Evidently one where we really
want a GUID on requests and really don't want it on other messages.

By contrast, the one certain use case for [message id], namely
correlation, can be supported by one rule:

   1. If a request contains a [message id], the reply must relate-to it.

To be clear, I'm open to supporting more use cases, and even to
strengthening the requirements for [message id], but I need to see
something considerably more concrete and specific than we've got now.

Yalcinalp, Umit wrote:

> 
>
>  
>
>>-----Original Message-----
>>From: Jonathan Marsh [mailto:jmarsh@microsoft.com] 
>>Sent: Tuesday, Jun 14, 2005 10:40 AM
>>To: tom@coastin.com; Yalcinalp, Umit
>>Cc: David Hull; public-ws-addressing@w3.org
>>Subject: RE: Why is [message id] required for requests but 
>>not for other messages?
>>
>>If the draft minutes are accurate, some of us voted to make 
>>[message ID]
>>mandatory, others to make it optional at all times, but the 
>>bulk of the
>>WG voted to simply 86 issue lc86.  I don't know what more to
>>productively say on the topic at this point.
>>    
>>
>
>I agree. In fact, that is exactly what I was trying to point out by my
>reply to David. We, at SAP, argued for making message ids mandatory. 
>
>It seems to me that we are rehashing the same thing again which we made
>a decision on. Lets move on. 
>
>Cheers, 
>
>--umit
>
>  
>
>>>-----Original Message-----
>>>From: public-ws-addressing-request@w3.org [mailto:public-ws-
>>>addressing-request@w3.org] On Behalf Of Tom Rutt
>>>Sent: Monday, June 13, 2005 4:21 PM
>>>To: Yalcinalp, Umit
>>>Cc: David Hull; public-ws-addressing@w3.org
>>>Subject: Re: Why is [message id] required for requests but not for
>>>other messages?
>>>
>>>
>>>check my concern below umit's comment
>>>
>>>Yalcinalp, Umit wrote:
>>>
>>>      
>>>
>>>>
>>>>        
>>>>
>>>>>-----Original Message-----
>>>>>From: public-ws-addressing-request@w3.org
>>>>>[mailto:public-ws-addressing-request@w3.org] On Behalf Of 
>>>>>          
>>>>>
>>David Hull
>>    
>>
>>>>>Sent: Monday, Jun 13, 2005 3:07 PM
>>>>>To: public-ws-addressing@w3.org
>>>>>Subject: Why is [message id] required for requests but not
>>>>>for other messages?
>>>>>
>>>>>
>>>>>If [message id] is to be leveraged for uses other than 
>>>>>          
>>>>>
>>correlation,
>>    
>>
>>>>>particularly duplicate elimination and security, wouldn't those
>>>>>considerations apply at least equally well to non request/reply
>>>>>interactions?  If not, what is the basis for requiring
>>>>>[message id] for
>>>>>requests but not for other types of message?
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>          
>>>>>
>>>>Well, if you recall from the f2f, we wanted to require message id
>>>>        
>>>>
>>>across
>>>      
>>>
>>>>the board :-)
>>>>
>>>>
>>>>
>>>>        
>>>>
>>>but those semantics are not there in the present document, 
>>>      
>>>
>>MesssageId
>>    
>>
>>>is
>>>only required when correlation of a reply is required.
>>>
>>>Tom Rutt
>>>
>>>      
>>>
>>>>--umit
>>>>
>>>>
>>>>
>>>>        
>>>>
>>>--
>>>----------------------------------------------------
>>>Tom Rutt	email: tom@coastin.com; trutt@us.fujitsu.com
>>>Tel: +1 732 801 5744          Fax: +1 732 774 5133
>>>
>>>
>>>      
>>>
>>    
>>
>
>  
>

Received on Tuesday, 14 June 2005 18:40:27 UTC