On 11/6/05, Peter J. Farrell <[EMAIL PROTECTED]> wrote:
> Yes, I just noticed about it returning an array - however, the
> pluginArray is already available in the pluginManager...might as well
> use it.  I'm still trying to get a comprehensive grasp of the code base.

Returning an array is inefficient because the array is copied (when
assigned to anything).

> Also, you'll want to include any 1.0.10 users of Mach-II (I guess you
> could call it legacy applications?) - so the method injection
> implementation is still probably the best way of doing it

Agreed. That's the approach I took with Model-Glue.

> I'm not sure about using a java vector for the autowire step - are the
> performance increases realistic or theoretical?

Sounds theoretical to me...

> Also, I'm not sure about CS and BlueDragon

I would be surprised if CS (and in particular the AOP stuff) worked on
BD. Tartan doesn't work on BD (well, it may work on 6.2.1 or whatever
the latest hotfix is) because BD doesn't implement some well-document
CFMX behavior.

Frameworks shouldn't care about compatibility with BD and Railo - BD
and Railo should (and do) care about compatibility with CFMX. New
Atlanta have been pretty good about fixing their bugs to become
CFMX-compatible and allow frameworks to run correctly so I wouldn't
worry about adding workarounds to the frameworks to support broken
implementations.
--
Sean A Corfield -- http://corfield.org/
Got frameworks?

"If you're not annoying somebody, you're not really alive."
-- Margaret Atwood

Reply via email to