Hi Wang Rui, I have gone ahead an refactored the versioning of serializers so that it now uses a _firstVersion and _lastVersion value, this allow is to support a version range. Reading your introductory post on this thread I believe this is what you intended in terms of functionality so believe this is really a fix rather than a change in intended functionality.
The macro usage that you introduced for this feature remains in place, and didn't need to change, so we still have the REMOVE_SERIALIZER() feature, and now that I understand the usage fully I realize there is no need for an extra complexities in the version specification that I originally was fretting about. The only questions I have left on this topic is what to do about the Version value itself - should it be tied to the SOVERSION as you have done right now? I'm not sure it should be, I'm more inclined towards a manually updated version number. Perhaps this would be good topic for open discussion. The other question I have is about whether we should be limiting the output of serializers when the version range doesn't support the current version being outputted. Thoughts? Robert. On Tue, Nov 9, 2010 at 11:58 AM, Robert Osfield <[email protected]> wrote: > I have looking at your versioning code again and have come to the > conclusion that the way you have implemented the versioning is that > you can either have a serializer active from a minimum version onwards > for ever, or have a serializer active from version 0 till a specified > version. Now this will catch a number of different usage cases, but > it won't allow us to specify that a feature should be active between > version range. > > I believe that a version range will be required, and while it will add > the need for an extra version variable in the ObjectWrapper and > Serializer the logic in the serializer selection will be far more > straight forward - no need to play tricks with negative numbers to get > a different meaning. > > I'm tempted to dive in a tweak your implementation to do this. Thoughts? > > Robert. > _______________________________________________ osg-submissions mailing list [email protected] http://lists.openscenegraph.org/listinfo.cgi/osg-submissions-openscenegraph.org
