On Jul 8, 2007, at 9:56 PM, Elliot Metsger wrote:
David H. DeWolf wrote:
1) Pluto should focus on the following 3 goals (in order and as
stated in it's mission):
- reference implementation
- easy portlet testing/development
- embeddable container for portal/webapp developers
<snip>
The question at hand (auto-deployment in pluto) is borderline
functionality to me. It specifically hits the second goal of
pluto (easy portlet testing/development), but could also be seen
as enterprise functionality.
I see it primarily as the second goal, but also as the third goal
as well; the idea is that embedders will get hot-deployment and
assembly for "free" by implementing relatively simple callbacks.
Assembly is really my focus. I'd like that to be abstracted away
from the portlet developer. Having portlet devs assembling wars is
damaging for the reasons at the bottom of [0].
Hot deployment is a natural consequence of assembly - if the
"container" is going to perform assembly at deployment time, it may
as well put the assembled war in a place where the servlet
container will pick it up.
I'm totally open to how this is accomplished, if not in Pluto than
in where?
I think it might be ok to provide this stuff in the pluto portal but
I don't think you should try to push it into the portlet container.
Here's a couple situations where I think it might be more of a
problem than a solution:
1. IIUC the last time I looked jetspeed 2 did not have a servlet-per-
portlet but rather a servlet-per-portlet-app. So the assembly is
going to be really different.
2. I've been planning for years now to write a portlet deployer for
geronimo that would directly read the portlet.xml and deal with it
appropriately: for the pluto portal set up the servlets for each
portlet directly, and for jetspeed configure the single servlet
appropriately. Here, xml munging would not be required.
3. Hot deployment is a terrible idea that IMO people use only because
many containers don't have usable deployment tools. Building yet
another semifunctional hot deployment implementation into yet another
product just spreads the misery further. (I do know that for some
reason not everyone agrees with me on this :-)
I think making portlet app deployment easier is a great goal but I
don't think xml munging or hot deployment are necessarily parts of a
widely applicable solution.
thanks
david jencks
[0] http://www.nabble.com/forum/ViewPost.jtp?post=11490584&framed=y
