W3C home > Mailing lists > Public > public-ws-policy@w3.org > March 2007

RE: 4414 [new issue] Versioning and Extensibility cleanup

From: Asir Vedamuthu <asirveda@microsoft.com>
Date: Wed, 28 Mar 2007 18:04:14 -0700
To: Maryann Hondo <mhondo@us.ibm.com>, "public-ws-policy@w3.org" <public-ws-policy@w3.org>
Message-ID: <C9BF0238EED3634BA1866AEF14C7A9E534E5A2DCAF@NA-EXMSG-C116.redmond.corp.microsoft.com>
As requested, here are the comments that we discussed today. Part 1 contains editorial changes and sounds good. We have few comments on Part 2.

a) the first policy alternative offered should NOT contain the EndOfLife policy assertion type

If not, when a provider signals EOL using the hypothetical EndOfLife assertion then the provider would break compatibility with requestors that rely on the first policy alternative. The WG discussed this point at length at the BEA F2F in January.

b) If Company-X adds the EndOfLife policy  assertion type

When the WG adopted Dave O's proposal [1], WG amended the proposal by adding the adjective 'hypothetical' to all occurrences of the 'EndOfLife assertion'.

c) Thus to provide the widest possible alternative compatibility Company-X should NOT have the ignorable attribute on the EndOfLife policy assertion in the very first alternative.

This statement contradicts the previous statement captured in point a) (see above). Do not know why.

[1] http://www.w3.org/Bugs/Public/show_bug.cgi?id=4270#c1

Regards,

Asir S Vedamuthu
Microsoft Corporation


From: public-ws-policy-request@w3.org [mailto:public-ws-policy-request@w3.org] On Behalf Of Maryann Hondo
Sent: Friday, March 23, 2007 8:40 AM
To: public-ws-policy@w3.org
Subject: [new issue] Versioning and Extensibility cleanup


http://www.w3.org/Bugs/Public/show_bug.cgi?id=4414<http://www.w3.org/Bugs/Public/show_bug.cgi?id=4374>


Title:  versioning and extensibility in primer needs cleanup

Description:  the language in the primer should be consistent in its use of "policy" and "policy expression".  " Policy assertion" design and best practice should be moved to the Guidelines document.

Justification: the guidelines is supposed to contain guidance on policy assertion authoring and the primer on policy expression use ....but the text mixes the two roles and is confusing.

Proposal:

Part 1-----------------------------------3.8 Extensibility and Versioning (after example 3-12)
From:

In this versioning scenario, policy can be used to represent current and advanced behaviors in a non-disruptive manner: no immediate changes to existing clients are required and these clients can smoothly migrate to new functionality when they choose to. This level of versioning support in policy enables the same class of versioning best practices built into WSDL constructs such as service, port and binding.

To:

In this versioning scenario, a policy expression with multiple alternatives can be used to represent current and advanced behaviors in a non-disruptive manner: no immediate changes to existing clients are required and these clients can smoothly migrate to new functionality when they choose to. This level of versioning support in a policy expression  enables the same class of versioning best practices built into WSDL constructs such as service, port and binding.

Part 2 -------------------------------------------------------3.8.1 Ignorable and Versioning
In this section, the expiry date and time are not a "domain expression" but a new assertion. The last sentence  should be in the Guidelines document.

From:

3.8.1 Ignorable and Versioning
One potential use of the wsp:Ignorable attribute is to mark versioning related information. One scenario is that a service will support a particular version of a service until a certain point in time. After that time, the service will not be supported. The expiry date and time of the service would be a domain expression, but it could be marked as ignorable. When an alternative does have an expiry, it is usually useful to convey this information to help smooth the versioning process.
....
In a scenario such as this, where an assertion type is used for ignorable information, the use of strict or lax mode and presence or absence of the assertion type in the first version are important decisions. If Company-X wishes clients to always be able to ignore the assertion, particularly those using strict intersection, the first policy alternative offered should contain the policy assertion type. If Company-X adds the policy assertion type to a subsequent alternative, then requesters using strict mode will not understand the assertion type and the alternative with the ignorable information will not be compatible with the older version of the alternative as per the intersection rules. Thus the widest possible alternative compatibility will be to have the ignorable policy assertion type in the very first version of the alternative. The second alternative shown also contains the hypothetical EndOfLife policy Assertion type. Because the actual value is not known, a value that is roughly infinitely in the future is used. A subsequent policy alternative could refine the value.


To:


3.8.1 Ignorable and Versioning
One potential use of the wsp:Ignorable attribute is to mark versioning related information. One scenario is that a service will support a particular version of a service until a certain point in time. After that time, the service will not be supported. The expiry date and time of the service would be a new policy assertion[see Guidelines document for best practices on defining new policy assertions and the use of the ignorable attribute] , with an ignorable attribute.
....
In a scenario such as this, Company-X is acting as both an assertion author and a policy expression author.

As a policy expression author, when an assertion type is used for ignorable information, the use of strict or lax mode and presence or absence of the assertion type in the first version are important decisions.

<sidebar--- this is another case where "best practices" are evident in the primer as opposed to the guidelines. Also, I'm not sure whether we agree as a group that it is a "best practice" to add a fictitious assertion to every policy alternative to anticipate a future need. I believe the text below is also a valid way to address intersection and compatability. >

<alternative to current text>
If Company-X wishes clients to always be able to use the original policy expression,  particularly those using strict intersection, the first policy alternative offered should NOT contain the EndOfLife policy assertion type.
If Company-X adds the EndOfLife policy  assertion type to a subsequent alternative, then requesters using strict mode will not understand the EndOfLife policy assertion type and the new alternative with the ignorable information will not be compatible with the older version of the alternative as per the intersection rules. Thus to provide the widest possible alternative compatibility Company-X should NOT have the ignorable attribute on the EndOfLife policy assertion in the very first alternative. The second alternative  contains the EndOfLife policy Assertion type with an ignorable attribute.
Received on Thursday, 29 March 2007 01:04:39 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 8 January 2008 14:20:48 GMT