Definitely need to flush out the solution of least surprise. I think that people will expect the system property to override web.xml - This would allow production servers to set the property and ensure that no matter what is deployed (accidents occur), they are *always* running in Production stage mode.
We need to make sure we are clear whether the context parameter overrides the system property, or vice versa. I would think the system property would take precedence, but I'm open to counter arguments.
At first thought, allowing web.xml to override the system property seems to allow greater flexibility, but in fact, it provides equal flexibility because you could simply leave your development stages at "Development" or whatever.. in web.xml, then disable the system-property when needed, allowing the same flexibility (as an inverse operation.)
I am in favor of the System property overriding web.xml. I don't think it makes sense otherwise.
--LBOn Tue, Jan 26, 2010 at 6:32 PM, Dan Allen <dan.j.allen@gmail.com> wrote:
On Tue, Jan 19, 2010 at 5:03 PM, Andy Schwartz <andyschwartz@oracle.com> wrote:
Ed Burns wrote:Okay, thanks for the update Ed. I wasn't aware that there had been push back from the Servlet EG on this.
On Mon, 18 Jan 2010 15:16:20 -0500, Andy Schwartz <andy.schwartz@oracle.com> said:
AS> https://javaserverfaces-spec-publicdev.java.net/issues/show_bug.cgi?id=499
AS> Can we add this to the MR list?
Yes, but I'd like to get agreement that we don't care about the Servlet
EG's opinion on the VM-wide nature of system properties.
Personally I don't understand the nature of the objection. In some case fine-grained (application-specific) control is desired. We have addressed this case via the context parameter. There seems to be general agreement in our EG that a system-level property would also be beneficial and in particular would improve the ease of use of this feature for development-time scenarios (ie. no need for a web.xml or JNDI config). Not sure why we need to choose one approach vs. the other. Both serve a purpose.
The system property seems like a very reasonable option to me. It matches well with "starting the server in debug mode" and I could even see the two being tied together in an IDE server control (tangential to this decision, of course). Just like with debug mode, this should be something that can be controlled without a change to application code.
We need to make sure we are clear whether the context parameter overrides the system property, or vice versa. I would think the system property would take precedence, but I'm open to counter arguments.
-Dan
--
Dan Allen
Senior Software Engineer, Red Hat | Author of Seam in Action
Registered Linux User #231597
http://mojavelinux.com
http://mojavelinux.com/seaminaction
http://www.google.com/profiles/dan.j.allen