Re: [equinox-dev][prov] JRE versioning

[James D Miles <jdmiles@xxxxxxxxxx>]

Grouping sounds good to me. It should probably be more general than handling a special case for JREs. There might be a case for other artifacts as well.

Maybe one could do quite a bit with just a group field. If IUs belong to the same group then they provide equivalent type functionality and can be considered (it might still not resolve). For my example:
- IU request jclDesktop but J2SE is installed.
- grouping is allowed
- both are identified as members of same equivalent group.
- allow J2SE to remain.
The governor could also play a part in accepting or declining a choice.

If we need a group IU then hopefully we don't have to know ahead of time which JRE IUs might be capable of attaching. The JRE IU should be able to declare itself a match.

David R Stevenson/Redmond/IBM@IBMUS

David R Stevenson/Redmond/IBM@IBMUS
Sent by: equinox-dev-bounces@xxxxxxxxxxx

10/11/2007 01:01 PM

Please respond to
Equinox development mailing list <equinox-dev@xxxxxxxxxxx>

To	Equinox development mailing list <equinox-dev@xxxxxxxxxxx>
cc
Subject	[equinox-dev][prov] JRE versioning

Another possibility for handing the multiple variants of JREs is to have a JRE (or java execution environment) group with selectable content so that the actual variant and version can be chosen via filters. You can think of this approach as corresponding to the way that arch, os, and ws are currently handled by eclipse. A bunch of details on exactly how to represent the myriad JREs and their relationships would have to be worked out.  Similarly we would have to craft very carefully the leaf IUs for jres to meet the use cases we imagine. 

If we did this, then our eclipse generator could use the execution environment information from the manifest to construct required capabilities on the appropriate jre - currently we ignore the execution environment and don't generate dependencies. 

        - Dave

James D Miles/Austin/IBM@IBMUS
Sent by: equinox-dev-bounces@xxxxxxxxxxx

10/11/2007 09:30 AM

Please respond to
Equinox development mailing list <equinox-dev@xxxxxxxxxxx>

To	equinox-dev@xxxxxxxxxxx
cc
Subject	[equinox-dev][prov]

>From Questions for M3 
"Should versions be pluggable (the current version does not properly reason about the version of the JRE"

JRE versions do go past the version definition that currently exists. I will only give you our experience using versions with the current update. Translate to the new terminology as needed.
We used the major number to create a hierarchy of different types of JREs. For instance we used 2.0.0 for j2se and 1.0.0 for jclDesktop. These were used for a parent feature. The child feature contained the actual JVM feature. This allowed us to provision JVMs like any other artifact. It allowed us to treat JVMs as equivalent from the ID perspective.

However it does not properly encapsulate the policy decisions that are necessary. If a platform initially specifies jcldesktop, version 1.0.0, then the user changes the JRE to J2SE, version 2.0.0, what gets done on an upgrade that specifies jclDesktop. So there likely needs to be another parameter/policy that specifies a user or platform choice (assuming either choice will resolve). In general updates only occur on versions. However IU with ID for jclDesktop does not equate to an ID for J2SE. The problem is how to know that they can replace each other.
_______________________________________________
equinox-dev mailing list
equinox-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/equinox-dev_______________________________________________
equinox-dev mailing list
equinox-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/equinox-dev