On Nov 28, 2005, at 1:41 AM, Manu George wrote:
Hi David,
Won't it be better to validate in both the cases to take
care of the unusual situations you mentioned? Is there any specific
reason why the validation is not done in the second case?
I don't see how we could do any meaningful validation. You can name
any gbean in any configuration anything you want, including giving it a
name that appears to be from a different configuration. AFAIK all the
names constructed from the "name" form are valid object/gbean names.
To see why we can't assume that the gbean referenced is available in
the server at deployment time, consider 2 ears A and B. A has some cmp
ejbs, however the datasource they use are in B, as is a web app that
uses the ejbs. If you deploy A first, the datasource is not available
for the cmp beans. If you deploy B first, the ejbs are not available
for the web app. One or the other can't use a resolution method that
requires the gbean to be present at the time of resolution
thanks
david jencks
Thanks
Manu
