On 6/9/06, Allen Gilliland <[EMAIL PROTECTED]> wrote:
comments inline ...
Dave Johnson wrote:
> Comments inline...
My plan would simply be to deprecate all of the old stuff in the context
rather than try and EOL it. We'd do that in these general steps ...
1. Round up the current context loading stuff and any objects it uses
and group it all together somewhere that is identified as legacy context
loading crud. This includes the current page model, page helper, etc.
Something like ui.rendering.legacy or ui.rendering.old.
2. Leave the legacy context loading process in place on the page
servlet, which is the only place that users have templates that use that
stuff. Remove it from everywhere else, feed, search, etc.
3. Define completely new page models to be used moving forward. These
models will be totally self contained and designed how we want things to
work moving foward.
4. Rewrite our existing velocity templates and themes to use only the
new page models and any other supporting objects that we want to
continue supporting for the long term. To lighten this burden I suggest
that we make 3.0 the point at which we EOL all the old themes and pick
just a couple new ones to refactor and ship with Roller for 3.0 and beyond.
5. Optionally we can define some kind of mechanism to turn off the old
context loading stuff. Probably the best way to do this is to begin
tracking what version of Roller the person began their installation
with. So starting with new installations 3.0 we could have the old
context loading stuff disabled by default. This is still a possible
problem though, if someone gets their hands on old templates/themes in a
3.0+ install they will no longer work, so we need to tread carefully here.
So the basic approach is to isolate all of the old stuff and consider it
deprecated. Then change our existing stuff to only use the new code
which we want to use moving forward.
The other fairly cool thing about this is that we would only document
the new models and objects that we really want to support as we create
them. We don't want to tell people about the old stuff and give them a
chance to start doing things the wrong way.
OK. I'm game.
I'm going to propose entirely new page models and macros. It doesn't
make sense to muck around with ContextLoader anymore if we're going
to EOL the old page models and macros in the 3.0 timeframe.
- Dave
