Re: Version attribute for WSDL

Jonathan,

Yup, not messing with the namespace itself for versioning (or not 
incorporating versioning into the namespace URI) is what I had been 
calling for (repeatedly :) in my follow-ups on this as well. People 
perhaps have been confused by the sequencing of the follow up messages. 
E.g. David's follow-up on my note (instead of perhaps William's), when 
we David and I both have the same view point.

Regards, Prasad

-------- Original Message --------
Subject: 	RE: Version attribute for WSDL
Resent-Date: 	Wed, 25 Feb 2004 13:48:30 -0500 (EST)
Resent-From: 	www-ws-desc@w3.org
Date: 	Wed, 25 Feb 2004 10:48:18 -0800
From: 	Jonathan Marsh <jmarsh@microsoft.com>
To: 	<www-ws-desc@w3.org>


David's point about not actually munging the namespace URI is important 
- the actual namespace URI may not be under the control of the author of 
the WSDL.  This is true even within the W3C - WGs can "suggest" that the 
W3C webmaster allocate a particular namespace URI but the webmaster has 
the ultimate say.  We don't want to introduce a spec that forces 
namespace URIs to be constructed in a way that violates the conventions 
a domain owner may have in effect.

 

------------------------------------------------------------------------

From: www-ws-desc-request@w3.org [mailto:www-ws-desc-request@w3.org] On 
Behalf Of David Orchard
Sent: Monday, February 23, 2004 6:32 PM
To: Prasad Yendluri; www-ws-desc@w3.org
Subject: RE: Version attribute for WSDL

 

What I had suggested before was that given an existing namespace 
(http://foo.com/myws/2), one could combine the ns and a version 
attribute into a URI to do differentiation in the following ways:

 

- Using "/" - thus the version URIs could be http://foo.com/myws/2/1 and 
http://foo.com/myws/2/1/1. 

- using ";" - they could be http://foo.com/myws/2;1 and 
http://foo.com/myws/2;1/1

- using "#" - http://foo.com/myws/2#1 and http://foo.com/myws/2#1/1

- using query params as defined by the HTML spec - 
http://foo.com/myws/2?version=2.1

 

In none of these cases would the URI be dereferencable, so there's no 
worry about the issue that the frag-id isn't sent to the server.

 

I personally tend to like the "/" as it reinforces the hierachical 
nature of the names.  I think we've had a lot of implied concern over 
creating URIs by adding things in the path, but I'm don't see the 
problem.   

 

Dave

    -----Original Message-----
    From: www-ws-desc-request@w3.org
    [mailto:www-ws-desc-request@w3.org]On Behalf Of Prasad Yendluri
    Sent: Monday, February 23, 2004 5:34 PM
    To: www-ws-desc@w3.org
    Subject: Re: Version attribute for WSDL

    Hi William,

    I agree this is a potential way to approach this, assuming client
    will look for special patterns in the namespace URI etc. I am not
    convinced however, moving everything to a new namespace for minor
    change is desirable. I see a namespace change to be a major one.

    Regards, Prasad

    ------- Original Message --------

    Subject:

    	

    RE: Version attribute for WSDL

    Date:

    	

    Mon, 23 Feb 2004 11:39:12 -0800

    From:

    	

    Vambenepe, William N <vbp@hp.com> <mailto:vbp@hp.com>

    To:

    	

    Prasad Yendluri <pyendluri@webmethods.com>
    <mailto:pyendluri@webmethods.com>, <www-ws-desc@w3.org>
    <mailto:www-ws-desc@w3.org>

     

    Not necessarily.

     

    If the consumer is aware of the rules we set up for building the
    "versioned" URI, it can examine the namespace URI and find out that
    this URI is not the one it is expecting but is one that is
    compatible with the one it is expecting.

     

    For example, if the client has been built based on a WSDL in namespace

    http://vambenepe.com/myWebService/version/2/1

    and it finds a service that implements a WSDL defined in namespace

    http://vambenepe.com/myWebService/version/2/1/1

    then, if the client understands our normative rules for creating
    versioned namespace URI and if these rules say that any WSDL
    construct defined in a namespace ending with version/2/1/1 must be
    backward compatible with the same constructs defined in the same
    namespace except ending with version/2/1, then the client knows is
    can invoke this service.

     

    If the client is unaware of these rules, then yes, it can only use
    services implementing WSDL descriptions in the exact same namespace.
    This is graceful degradation.

     

    The same semantic (about major/minor changes, forward/backward
    compatibility, etc...) that you can express in a "version" attribute
    can be expressed using rules on namespace URI. WSDL consumers that
    would support the "version" attribute could just as easily support
    these rules as applied to the namespace URI. And for those that
    don't want to be bothered with this support for versioning, putting
    the info in the namespace URI allows them to keep working the way
    they do today. While with the version attribute you force people to
    either support it or take the risk of nasty stuff happening.

     

    William

        -----Original Message-----
        From: www-ws-desc-request@w3.org
        <mailto:www-ws-desc-request@w3.org>
        [mailto:www-ws-desc-request@w3.org] On Behalf Of Prasad Yendluri
        Sent: Friday, February 20, 2004 2:38 PM
        To: www-ws-desc@w3.org <mailto:www-ws-desc@w3.org>
        Subject: Re: Version attribute for WSDL

        William,

        The difficulty is that backward compatible versioning is not
        possible with this.
        That is, applications that worked with the earlier version no
        longer work this version,
        as this is a "different interface" rather than a minor version
        of one.

        This approach is always available for changes that have no
        backward compatibility baggage associated but,
        I believe the impetus behind the solution being sought is the
        backward compatibility aspect, where vendors can make
        changes without having to worry about impacting current users
        and not having required to publish an entirely
        new WSDL in new a namepace each time a minor revision is needed etc.

        Regards, Prasad

        -------- Original Message --------

        Subject:

        	

        RE: Version attribute for WSDL

        Resent-Date:

        	

        Thu, 19 Feb 2004 22:53:45 -0500 (EST)

        Resent-From:

        	

        www-ws-desc@w3.org <mailto:www-ws-desc@w3.org>

        Date:

        	

        Thu, 19 Feb 2004 19:53:33 -0800

        From:

        	

        Vambenepe, William N <vbp@hp.com> <mailto:vbp@hp.com>

        To:

        	

        Prasad Yendluri <pyendluri@webmethods.com>
        <mailto:pyendluri@webmethods.com>, <www-ws-desc@w3.org>
        <mailto:www-ws-desc@w3.org>

         

This is indeed what I am asking. What problems do you see? A different

interface (no matter how different) gets a different QName. So we don't

mess up anything and start to see interactions fail that we thought

should work, or requests having a different effect from what was

expected. This is a difficult area, let's play it self. Having

established this safe foundation, we can now look at ways to offer

additional features when the interfaces are indeed versions of one

another, by having a convention on how to build a the namespace URI of a

new interface based on the namespace URI of the older version. If people

want to take advantage of this convention, great. If they don't then

they don't have to worry about it and their code won't suddenly break

because they forgot to check the "version" and follow whatever rules we

define about minor/major versions.

 

What I don't like about the approach that uses a new "version" attribute

is that it puts additional constraints and a heavy burden on people who

don't care about versioning. The approach I propose allows people who

don't care about expressing versioning to ignore it and be safe.

 

William

 

> -----Original Message-----

> From: www-ws-desc-request@w3.org <mailto:www-ws-desc-request@w3.org> 

> [mailto:www-ws-desc-request@w3.org] On Behalf Of Prasad Yendluri

> Sent: Thursday, February 19, 2004 2:24 PM

> To: www-ws-desc@w3.org <mailto:www-ws-desc@w3.org>

> Subject: Re: Version attribute for WSDL

> 

> 

> 

> William,

> 

> >Prasad, I don't see what the problem is with creating a new 

> namespace URI by using a >previous namespace URI as a base.

> 

> There is no issue with creating a new URI that represents a version with 

> the namespace URI as the base. However if you are asking that we make 

> that the new namespace URI (@targetNamespace on the definitions element) 

> of the versioned WSDL effectively using the namespace attribute to 

> version,  I do see a problem. I was not sure however if  that is what 

> you are proposing..

> 

> Regards, Prasad

> 

> > -----Original Message----- 

> > From: www-ws-desc-request@w3.org <mailto:www-ws-desc-request@w3.org> on behalf of 

> Vambenepe, William N 

> > Sent: Thu 2/19/2004 11:28 AM 

> > To: Prasad Yendluri; www-ws-desc@w3.org <mailto:www-ws-desc@w3.org> 

> > Cc: 

> > Subject: RE: Version attribute for WSDL

> > 

> > 

> > David, in your proposal, you suggest to create a URI 

> based on the namespace URI of the "older" interface. That 

> sounds good, but where does this URI go? First I assumed you 

> meant it goes in an attribute of type anyURI that we would 

> add to interface, and I replied to you that this URI should 

> go as the namespace URI for the "newer" interface instead. 

> Then I re-read your text and I thought maybe this is what you 

> meant all along. Now I am confused again. Can you please 

> clarify where in the WSDL you would put the "versioned" URI? 

> As the namespace of the "newer" interface or not.

> >  

> > Prasad, I don't see what the problem is with creating a 

> new namespace URI by using a previous namespace URI as a base.

> >  

> > William

> >  

> >

> >        -----Original Message-----

> >        From: www-ws-desc-request@w3.org <mailto:www-ws-desc-request@w3.org> 

> [mailto:www-ws-desc-request@w3.org] On Behalf Of Prasad Yendluri

> >        Sent: Tuesday, February 17, 2004 5:31 PM

> >        To: www-ws-desc@w3.org <mailto:www-ws-desc@w3.org>

> >        Subject: Re: Version attribute for WSDL

> >        

> >        

> >

> >

> >        David Orchard wrote:

> >        

> >

> >                Creating a URI from the namespace name 

> and version is not messing with the namespace URI. 

> >                

> >

> >        I did not suggest that at all. My response is 

> based on William's note below, the text around "Yes, I fully 

> agree with doing this in the URI that represents the 

> namespace." I interpreted that to mean changing the namespace 

> URI. If we are not, I have no issue (on that).

> >        

> >        

> >

> >                It's about having a normative way of 

> generating a URI from a base URI, in this case an NS URI, and 

> a secondary resource identifier, in this case a version 

> identifier.  We could use #, ;, / to separate.  Just depends 

> on whether you want the server to see the secondary resource 

> identifier or not.  This seems like a new symbol space to me...

> >                 

> >                This is all completely valid and 

> expected within the web and web services architecture..

> >                 

> >                Dave

> >                -----Original Message-----

> >                From: www-ws-desc-request@w3.org <mailto:www-ws-desc-request@w3.org> 

> [mailto:www-ws-desc-request@w3.org]On Behalf Of Prasad Yendluri

> >                Sent: Tuesday, February 17, 2004 4:53 PM

> >                To: www-ws-desc@w3.org <mailto:www-ws-desc@w3.org>

> >                Subject: Re: Version attribute for WSDL

> >                

> >                

> >

> >                        Messing with the namespace URI 

> (and effectively changing namespace) for versioning seems 

> undesirable to me. 

> >                        I like to keep version a 

> separate attribute distinct from the namespace and actually I 

> prefer that it be a +ve numerical.

> >                        A higher value representing a 

> younger (subsequent) version than the lower valued one. Also IMO we 

> >                        must clearly define the 

> semantics of this attribute lest we end up creating another 

> area of confusion. 

> >                        The following semantics look good to me.

> >                        

> >                        The version attribute indicates 

> a "minor" revision of the definition or

> >                        interface. Specifically, a 

> "minor" revision indicates that a client using

> >                        a WSDL with a version attribute 

> less-than the current value is expected

> >                        to continue to function.

> >                        In essence I am for Tom's (2nd) 

> proposal below except the approach suggested in example 2. 

> When an incompatible change is made, 

> >                        it should be left up to the 

> WSDL writer to decide how to change the namespace URI rather 

> the spec dictating it.

> >                        

> >                        I don't quite understand the 

> issue with defaulting the version to "1" if not set 

> explicitly however...

> >                        

> >                        Regards, Prasad

> >                       

> >                        -------- Original Message -------- 

> >Subject:        RE: Version attribute for WSDL 

> >Resent-Date:    Fri, 13 Feb 2004 18:49:15 -0500 (EST) 

> >Resent-From:    www-ws-desc@w3.org <mailto:www-ws-desc@w3.org>     

> >Date:   Fri, 13 Feb 2004 15:48:53 -0800       

> >From:   Vambenepe, William N <vbp@hp.com> <mailto:vbp@hp.com> <mailto:vbp@hp.com>        

> >To:     David Orchard <dorchard@bea.com> <mailto:dorchard@bea.com> 

> <mailto:dorchard@bea.com> , <paul.downey@bt.com> <mailto:paul.downey@bt.com> 

> <mailto:paul.downey@bt.com> , <tomj@macromedia.com> <mailto:tomj@macromedia.com> 

> <mailto:tomj@macromedia.com> , <www-ws-desc@w3.org> <mailto:www-ws-desc@w3.org> 

> <mailto:www-ws-desc@w3.org>     

> >

> >                        My mistake, I thought you were 

> suggesting doing URI tricks inside a

> >                        "version" attribute that would 

> be of type URI. Yes, I fully agree with

> >                        doing this in the URI that 

> represents the namespace. This is what I

> >                        meant in a previous email: 

> "maybe what we need instead is an optional

> >                        convention on how to build 

> interface QNames that convey versioning

> >                        information. Whether that 

> convention belongs in the WSDL spec is another

> >                        question...""

> >                        

> >                        So +1 from me that this is the 

> right approach. I am not sure this group

> >                        needs to specify that, but it's 

> fine by me if many people want it in our

> >                        spec.

> >                        

> >                        William

> >                        

> >                        > -----Original Message-----

> >                        > From: www-ws-desc-request@w3.org <mailto:www-ws-desc-request@w3.org> 

> >                        > 

> [mailto:www-ws-desc-request@w3.org] On Behalf Of Vambenepe, William N

> >                        > Sent: Friday, February 13, 

> 2004 12:41 PM

> >                        > To: David Orchard; 

> paul.downey@bt.com <mailto:paul.downey@bt.com>; tomj@macromedia.com <mailto:tomj@macromedia.com>; 

> >                        > www-ws-desc@w3.org <mailto:www-ws-desc@w3.org>

> >                        > Subject: RE: Version 

> attribute for WSDL

> >                        > 

> >                        > 

> >                        > > I wonder if we could play 

> some magic trick and say that the 

> >                        > > minor version is a relative 

> URI from the namespace name, and 

> >                        > > then the "match" could be 

> of the strings.  A nice use of URIs 

> >                        > > for comparison imo.

> >                        > 

> >                        > Why not play that trick on 

> the URI part of the QName of the interface?

> >                        > 

> >                        > William

> >                        > 

> >                        > > -----Original Message-----

> >                        > > From: www-ws-desc-request@w3.org <mailto:www-ws-desc-request@w3.org> 

> >                        > > 

> [mailto:www-ws-desc-request@w3.org] On Behalf Of David Orchard

> >                        > > Sent: Friday, February 13, 

> 2004 12:13 PM

> >                        > > To: paul.downey@bt.com <mailto:paul.downey@bt.com>; 

> tomj@macromedia.com <mailto:tomj@macromedia.com>; www-ws-desc@w3.org <mailto:www-ws-desc@w3.org>

> >                        > > Subject: RE: Version 

> attribute for WSDL

> >                        > > 

> >                        > > 

> >                        > > Ah Paul,

> >                        > > 

> >                        > > I had earlier thought about 

> using URIs for the "minor" 

> >                        > > version # and the problem 

> of multiple nested versions and you 

> >                        > > are probably right about 

> the problem of increasing minor versions.  

> >                        > > 

> >                        > > Tell me though, is 3.3 

> compatible with 3.2.1.1?  I would 

> >                        > > assume they would have to be.

> >                        > > 

> >                        > > I wonder if we could play 

> some magic trick and say that the 

> >                        > > minor version is a relative 

> URI from the namespace name, and 

> >                        > > then the "match" could be 

> of the strings.  A nice use of URIs 

> >                        > > for comparison imo.

> >                        > > 

> >                        > > Dave

> >                        > > 

> >                        > > > -----Original Message-----

> >                        > > > From: www-ws-desc-request@w3.org <mailto:www-ws-desc-request@w3.org> 

> >                        > > 

> [mailto:www-ws-desc-request@w3.org]On

> >                        > > > Behalf Of paul.downey@bt.com <mailto:paul.downey@bt.com>

> >                        > > > Sent: Friday, February 

> 13, 2004 10:02 AM

> >                        > > > To: dorchard@bea.com <mailto:dorchard@bea.com>; 

> tomj@macromedia.com <mailto:tomj@macromedia.com>; www-ws-desc@w3.org <mailto:www-ws-desc@w3.org>

> >                        > > > Subject: RE: Version 

> attribute for WSDL

> >                        > > > 

> >                        > > > 

> >                        > > > 

> >                        > > > I like this too, 

> especially the defaulting on extension.

> >                        > > > 

> >                        > > > My small concern is using 

> the integer to indicate the relationship

> >                        > > > between versions 

> precludes branches, unless we allowed a  SCCS/RCS/CVS 

> >                        > > > style numbering system, e.g:

> >                        > > > 

> >                        > > > 1 -> 2 -> 3 -> 4 -> 5 

> >                        > > >           |

> >                        > > >           +-> 3.1 -> 3.2 -> 3.3 

> >                        > > >                      |

> >                        > > >                      +-> 3.2.1

> >                        > > >                          |

> >                        > > >                          

> +-> 3.2.1.1 

> >                        > > > 

> >                        > > > i imagined the proper W3C 

> way would be to use a URI for the 

> >                        > > > version and 

> >                        > > > relate them using syllogisms ?

> >                        > > > 

> >                        > > > Paul

> >                        > > > 

> >                        > > > 

> >                        > > > -----Original Message-----

> >                        > > > From: www-ws-desc-request@w3.org <mailto:www-ws-desc-request@w3.org> 

> >                        > > 

> [mailto:www-ws-desc-request@w3.org]On Behalf Of David Orchard

> >                        > > > Sent: 13 February 2004 17:47

> >                        > > > To: 'Tom Jordahl'; 

> www-ws-desc@w3.org <mailto:www-ws-desc@w3.org>

> >                        > > > Subject: RE: Version 

> attribute for WSDL

> >                        > > > 

> >                        > > > 

> >                        > > > I like this as a 

> strawman.  And the idea of not inheriting the version

> >                        > > > attribute makes a certain 

> sense too, as it requires the "extender" to make a

> >                        > > > conscious decision.  

> Though defaulting to "1" does have the problem that the

> >                        > > > extender might not be 

> compatible.  If there were some way in the "extension"

> >                        > > > of knowing that the 

> extensions could be ignored, then "1" makes sense.

> >                        > > > 

> >                        > > > cheers,

> >                        > > > Dave

> >                        > > > 

> >                        > > > > -----Original Message-----

> >                        > > > > From: 

> www-ws-desc-request@w3.org <mailto:www-ws-desc-request@w3.org> 

> >                        > > > 

> [mailto:www-ws-desc-request@w3.org]On Behalf Of Tom Jordahl

> >                        > > > > Sent: Friday, February 

> 13, 2004 7:06 AM

> >                        > > > > To: 'www-ws-desc@w3.org <mailto:www-ws-desc@w3.org>'

> >                        > > > > Subject: RE: Version 

> attribute for WSDL

> >                        > > > >

> >                        > > > >

> >                        > > > > I guess I understand 

> the desire to have "real" versioning support in WSDL

> >                        > > > > 2.0.  I do too. But my 

> proposal came out of the F2F, where we had a long,

> >                        > > > > and I believe 

> fruitless, discussion about all of this.

> >                        > > > >

> >                        > > > > I do not believe we can 

> have a section in our specification about versioning

> >                        > > > > and say "if you want 

> versioning, change the namespace". With a small

> >                        > > > > addition to the syntax, 

> we can give users some help in managing change in

> >                        > > > > their web services.

> >                        > > > >

> >                        > > > > I am willing to apply 

> semantics to the version attribute if this group

> >                        > > > > thinks that they can 

> move forward in a productive way.  How about these

> >                        > > > > changes as a straw man 

> for discussion:

> >                        > > > >

> >                        > > > >  - The version 

> attribute is part of the infoset (a.k.a. the component model)

> >                        > > > >

> >                        > > > >  - The version 

> attribute has type xsd:positiveInteger

> >                        > > > >

> >                        > > > >  - The version 

> attribute has a default value of 1.

> >                        > > > >

> >                        > > > >  - The version 

> attribute indicates a "minor" revision of the definition or

> >                        > > > >    interface. 

> Specifically, a "minor" revision indicates that a client using

> >                        > > > >    a WSDL with a 

> version attribute less-than the current value is expected

> >                        > > > >    to continue to function.

> >                        > > > >

> >                        > > > >  - When an interface 

> extends another interface, the version attribute of the

> >                        > > > >    interface is NOT 

> inherited - it must be explicitly set on the interface,

> >                        > > > >    and if is not, the 

> interface has the default version attribute (1).

> >                        > > > >

> >                        > > > > Example 1: Version 1 of 

> my interface has two operations. I release a new

> >                        > > > > WSDL that adds a third 

> operation, and change the version attribute to 2.

> >                        > > > > Clients who are using 

> the previous version of the WSDL continue to function.

> >                        > > > >

> >                        > > > > Example 2: My WSDL has 

> a purchase order type defined and a target namespace

> >                        > > > > of 

> "http://example.org/myservice" <http://example.org/myservice> > <http://example.org/myservice> 

> .  I change my purchase order 

> to include

> >                        > > > > several new elements 

> and rename some of the old ones.  Since this change

> >                        > > > > will break 

> compatibility, I change the target namespace to

> >                        > > > > 

> "http://example.org/myservice/v2" <http://example.org/myservice/v2> 

<http://example.org/myservice/v2> .  My service can now easily tell the

>                          > > > > difference between clients that

are using the original WSDL

>                          > > > > and the new one.

>                          > > > >

>                          > > > >

>                          > > > > --

>                          > > > > Tom Jordahl

>                          > > > > Macromedia Server Development

>                          > > > >

>   

     

Received on Wednesday, 25 February 2004 14:19:38 UTC