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

Mark Little wrote:

>
>
MessageID can be in the message even if a reply is not expected.  If a 
reply is expected then the MessageID must be present.  This is the status
quo behaviour of the Last Call spec.

Tom Rutt

>
> Conor P. Cahill wrote:
>
>> Mark Little wrote on 6/16/2005, 9:13 AM:
>>
>> >
>> > I didn't mean to imply you'd said sessions explicitly and thought the
>> > rest of my message made that clear. It's just that the term 
>> correlation
>> > id is often used when talking about sessions. If you're just talking
>> > about simply tying together a request and a response (with subsequent
>> > requests having different "ids") then I reiterate that I don't have a
>> > problem with MessageID, or (going back to the mid 80's when RPCs were
>> > the king) SequenceNumber. I think shifting to CorrelationID runs the
>> > risk of increasing the confusion you mention.
>>
>> So, to summarize, I'm saying that MessageID has proven to be *extremely*
>> confusing to everybody, incuding most of the people in this group.  You
>> are saying that choosing the name CorrelationID may also have some level
>> of confusing.
>>  
>>
> Yes, that's a fair summary.
>
>> So, how about using RequestID.
>>  
>>
> I think it more closely maps to the requirements, particularly since 
> you can't have a MessageID/RequestID without a ReplyTo. However, what 
> are the semantics if you have a RequestID and no ReplyTo? Doesn't the 
> syntax of RequestID imply a response is also required and hence the 
> name might still be confusing? (Just playing Devil's Advocate.)
>
> Mark.
>
>> Conor
>>
>>
>>  
>>
>
>


-- 
----------------------------------------------------
Tom Rutt	email: tom@coastin.com; trutt@us.fujitsu.com
Tel: +1 732 801 5744          Fax: +1 732 774 5133

Received on Thursday, 16 June 2005 13:55:27 UTC