W3C home > Mailing lists > Public > public-ws-addressing@w3.org > November 2004

Re: Mandator wsa:Action (was Re: WS-Addr issues)

From: Mark Little <mark.little@arjuna.com>
Date: Fri, 5 Nov 2004 08:36:09 +0000
Message-Id: <BDEA08B8-2F05-11D9-84E9-00039399DACE@arjuna.com>
Cc: "Francisco Curbera" <curbera@us.ibm.com>, <public-ws-addressing-request@w3.org>, <public-ws-addressing@w3.org>
To: "David Orchard" <dorchard@bea.com>

David, apologies for the delay in replying and I see that the 
discussion has moved on since last night (my time). Nevertheless I'll 
respond inline.

On 4 Nov 2004, at 17:56, David Orchard wrote:

> Mark,
>
> I think you misunderstand me.  I'm not saying design architectures
> around bad toolkits.  I am saying architecture will guide component
> design.  that I'm saying by making something like Action optional, 
> there
> is a natural and predictable way that people will build toolkits.

Understood. However, there are a couple of points I need to make:

(i) the financial saying of "past performance is no indicator of 
current or future performance" comes into play. Just because you've 
seen this happen before doesn't mean it will happen again. If there 
really is a need for this in some applications, then people will use 
it.

(ii) building on (i), as I've said before, I've seen quite a few 
implementations that just see wsa:Action as baggage they have to fill 
in but don't use. I think we should be looking to make mandatory only 
those elements that are used in every situation. If you can show me 
that that is the case for wsa:Action then I'd be interested. This model 
of mandatory elements only for global requirements works for other 
specifications, so I don't see why it won't work for us.

>
> For example, the Web is based upon self-describing messages.  They are
> self-describing because the media-type of the representation is 
> included
> the message.  This enables software to correctly dispatch n different
> representation types that are retrieved as part of a given
> operaion/action.  You could argue that the media-type should have been
> optional for something like jpeg/gif and that any XML content type 
> could
> be determined by looking at the body.  Oh, but wait, there may be some
> xml vocabularies, like XSLT and XENC, that can be embedded in others.
> They are the media type but they allow an "embedded" mode where they
> aren't the doc root.  So the "body" child isn't it.  By enforcing
> mandatory media-type, a whole bunch of web arch problems where
> prevented.

I don't see this as the same thing at all. Firstly, MIME types can be 
strongly argued as required on all interactions. Secondly, this isn't 
addressing, it's dispatch. I think it's simply the wrong place in the 
architecture to put this (if it needs defining at all).

>
> As for the meaning of addressing, which is what I think you are tilting
> at, I think Action is about how a resource is addressed as much as a
> URI.  There is a big distinction between our informal and not written
> down definition of what "addressing" is and a formal definition of an
> identifier like URIs.  In my world of what "addressing" means, the
> Action of a message will disambiguate between 2 equal message body
> contents, the point that Paco made ealier and I +1ed.

We'll have to disagree then about what is and isn't addressing in this 
case. However, I think if you look at other distributed architectures 
over the past 30 years you'll be in the minority ;-)

>
> I'd rather not get into the age old computing discussion about whether
> opcodes are data or not.  Please, let's not go there.

Agreed.

Mark.

>
> Dave
>
>> -----Original Message-----
>> From: Mark Little [mailto:mark.little@arjuna.com]
>> Sent: Thursday, November 04, 2004 9:39 AM
>> To: David Orchard; Francisco Curbera
>> Cc: public-ws-addressing@w3.org; public-ws-addressing-request@w3.org
>> Subject: Re: Mandator wsa:Action (was Re: WS-Addr issues)
>>
>>> The real problem is the same problem we had with the optional soap
> 1.1
>>> action http header.  Software can't count on it being there, so they
> end
>>> up looking inside the body as "the one true and certified source of
>>> action" which effectively pushed everybody into RPC land.  This
> happened
>>> because all the toolkits had to support at least looking in the body
> and
>>> then not all did the look at action and thus the world was a worse
>>> place.
>>
>> That's a problem with the toolkits though. If wsa:Action is there, use
> it.
>> If it isn't, look in the body. To say that you always look in the body
>> because of poor toolkits is the wrong reason to forge an architecture.
> If
>> that were the main rule for anything, we'd still be working with
> TCP/IP
>> and
>> sendmsg/recvmsg.
>>
>>> I predict that an optional WSA:Action will have the same effect IF
> there
>>> is no mandatory/normative way of generating a WSA:Action infset
> property
>>> from any binding that hasn't serialized the WSA:Action as a soap
> header
>>> block.
>>>
>>> I don't want to live in the message bodies always contain the verb
> world
>>> any more.
>>
>> I understand. My preference would to be remove this from WS-Addr
> entirely
>> and put this into another spec. If we can't do that (because for some
>> reason
>> people really believe this is something to do with addressing), then
> let's
>> make it optional and assume that vendors can provide toolkits that
> work :-
>> )
>>
>> Mark.
>>
>>>
>>> Dave
>>>
>>>> -----Original Message-----
>>>> From: Mark Little [mailto:mark.little@arjuna.com]
>>>> Sent: Thursday, November 04, 2004 9:24 AM
>>>> To: David Orchard; Francisco Curbera
>>>> Cc: public-ws-addressing@w3.org;
> public-ws-addressing-request@w3.org
>>>> Subject: Mandator wsa:Action (was Re: WS-Addr issues)
>>>>
>>>> David, I changed the subject line - you're right in that regard.
>>>>
>>>> As for keeping wsa:Action mandatory, I think you're wrong ;-)
>>>>
>>>> What is the real problem with making this optional? What would
> break
>>> as a
>>>> result?
>>>>
>>>> Mark.
>>>>
>>>> ----
>>>> Mark Little,
>>>> Chief Architect,
>>>> Arjuna Technologies Ltd.
>>>>
>>>> www.arjuna.com
>>>>
>>>> ----- Original Message -----
>>>> From: "David Orchard" <dorchard@bea.com>
>>>> To: "Francisco Curbera" <curbera@us.ibm.com>; "Mark Little"
>>>> <mark.little@arjuna.com>
>>>> Cc: <public-ws-addressing@w3.org>;
>>> <public-ws-addressing-request@w3.org>
>>>> Sent: Thursday, November 04, 2004 4:40 PM
>>>> Subject: RE: WS-Addr issues
>>>>
>>>>
>>>>> +1.
>>>>>
>>>>> Arguing against action is like arguing against HTTP operations.
>>> Having
>>>>> one spot for Action will give all WS-A applications a much
> simpler
>>>>> processing model and enable a doc/literal world.
>>>>>
>>>>> Separately, can we pick better subject lines and focus the
>>> conversation
>>>>> a bit?  I think this thread is on mandatory Action.  I expect we
> are
>>>>> going to debate every single component's mandatory/optional
> nature
>>> and
>>>>> separating them would help a lot.
>>>>>
>>>>> Dave
>>>>>
>>>>>> -----Original Message-----
>>>>>> From: public-ws-addressing-request@w3.org
>>>>> [mailto:public-ws-addressing-
>>>>>> request@w3.org] On Behalf Of Francisco Curbera
>>>>>> Sent: Thursday, November 04, 2004 6:26 AM
>>>>>> To: Mark Little
>>>>>> Cc: public-ws-addressing@w3.org;
>>> public-ws-addressing-request@w3.org
>>>>>> Subject: Re: WS-Addr issues
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> The idea that the intent of the message is *always* embedded
> in
>>> the
>>>>> body
>>>>>> of
>>>>>> the message smells like SOAP-RPC in sheep clothes to me. I am
> not
>>>>> saying
>>>>>> that will never be the case, but you need to allow for the
> case in
>>>>> which
>>>>>> the same document type is used in different interactions - for
>>>>> example, a
>>>>>> customerInfo document could be sent as input to both an
> "update"
>>> and a
>>>>>> "create" operations.This "document centric" model is actually
> very
>>>>>> frequent
>>>>>> (it is no uncommon in CICS applications for example). To
> support
>>> this
>>>>>> model
>>>>>> you need either an Action header or something functionally
>>> equivalent.
>>>>>>
>>>>>> Paco
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>                       "Mark Little"
>>>>>>                       <mark.little@arjuna.com>        To:
>>>>> "Sanjiva
>>>>>> Weerawarana" <sanjiva@watson.ibm.com>,
>>> <public-ws-addressing@w3.org>
>>>>>>                       Sent by:                        cc:
>>>>>>                       public-ws-addressing-req        Subject:
>>> Re:
>>>>> WS-
>>>>>> Addr issues
>>>>>>                       uest@w3.org
>>>>>>
>>>>>>
>>>>>>                       11/04/2004 05:05 AM
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> Hi Sanjiva. Although not an answer to your question, I think
> it's
>>>>> worth
>>>>>> bringing up generally: personally I think wsa:Action should be
>>> dropped
>>>>> or
>>>>>> made optional. Why have an "op code" (which is essentially
> what it
>>> is)
>>>>>> embedded in an address? I can see that there are optimizations
>>> that
>>>>> could
>>>>>> be made to dispatching directly on the Action rather than
> having
>>> to
>>>>> parse
>>>>>> the body, but surely that's an implementation specific issue?
> I'd
>>> be
>>>>>> interested in knowing how many users of WS-Addressing actually
> use
>>>>> this
>>>>>> versus those that ignore it.
>>>>>>
>>>>>> Mark.
>>>>>>
>>>>>> ----
>>>>>> Mark Little,
>>>>>> Chief Architect,
>>>>>> Arjuna Technologies Ltd.
>>>>>>
>>>>>> www.arjuna.com
>>>>>> ----- Original Message -----
>>>>>> From: Sanjiva Weerawarana
>>>>>> To: public-ws-addressing@w3.org
>>>>>> Sent: Wednesday, November 03, 2004 7:42 PM
>>>>>> Subject: Re: WS-Addr issues
>>>>>>
>>>>>> Hi Steve,
>>>>>>
>>>>>> What's your view of dispatching with wsa:Action? Since those
> are
>>>>> required
>>>>>> to be unique that gives enough info to find the operation to
>>> dispatch
>>>>>> to within a service. The service itself is of course
> identified
>>> from
>>>>>> the <To> somehow.
>>>>>>
>>>>>> Sanjiva.
>>>>>>  ----- Original Message -----
>>>>>>  From: Vinoski, Stephen
>>>>>>  To: Doug Davis
>>>>>>  Cc: public-ws-addressing@w3.org
>>>>>>  Sent: Thursday, November 04, 2004 12:58 AM
>>>>>>  Subject: RE: WS-Addr issues
>>>>>>
>>>>>>  +1 to having a pointer to the WSDL itself in the EPR. We have
>>> found
>>>>> in
>>>>>>  working with our customers that having access to the service
>>>>> definition
>>>>>> is
>>>>>>  critical for applications that rely on pure dynamic
> dispatching.
>>>>>>
>>>>>>  --steve
>>>>>>        -----Original Message-----
>>>>>>        From: Doug Davis [mailto:dug@us.ibm.com]
>>>>>>        Sent: Wednesday, November 03, 2004 11:02 AM
>>>>>>        To: public-ws-addressing@w3.org
>>>>>>        Subject: WS-Addr issues
>>>>>>
>>>>>>
>>>>>>        I might have missed a formal request for "issues" from
> the
>>>>> public
>>>>>>        but since it appears there is now an issues list I
> thought
>>> I'd
>>>>> make
>>>>>>        some suggestions on possible issues for the WG's
>>> consideration:
>>>>>>
>>>>>>        issue: EPRs have WSDL bits - e.g. PortType,
> ServiceName.
>>> But
>>>>> no
>>>>>>        pointer to the actual WSDL itself - why not?  W/o the
> WSDL
>>> do
>>>>> these
>>>>>>        values mean anything?  And if we assume the consumer of
> the
>>> EPR
>>>>> has
>>>>>>        the WSDL why can't we assume they know the PortType and
>>>>>> ServiceName?
>>>>>>        Perhaps an example of how this would be used would
> clarify
>>> it
>>>>> for
>>>>>>        me.
>>>>>>
>>>>>>        issue: If a response message is expected then a
> wsa:ReplyTo
>>>>> MUST be
>>>>>>        included.  Does the absence of a wsa:ReplyTo imply a
>>> one-way
>>>>>>        message?  The spec seems to come very close to saying
> that.
>>>>> And
>>>>>>        does the presence of wsa:ReplyTo imply a two-way
> message?
>>> My
>>>>>>        preference would be to have a clear statement so that
> upon
>>>>>>        inspection of the message itself a processor can know
> if
>>> its a
>>>>>>        one-way or two-way w/o having to go back to the wsdl.
>>>>>>
>>>>>>        issue: wsa:FaultTo:  "This property may be absent if
> the
>>> sender
>>>>>>        cannot receive fault messages (e.g. is a one-way
>>> application
>>>>>>        message)."  But it also says that in the absence of
>>> wsa:FaultTo
>>>>> the
>>>>>>        wsa:ReplyTo/From may be used.  So, how does a client
> really
>>> say
>>>>>> that
>>>>>>        it doesn't want ANY fault messages at all but still be
>>> allowed
>>>>> to
>>>>>>        specify a wsa:From?
>>>>>>
>>>>>>        thanks
>>>>>>        -Doug
>>>>>>
>>>>>
>>>>>
>>>
>>>
>
Received on Friday, 5 November 2004 08:37:27 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 2 June 2009 18:34:59 GMT