Luciano Resende wrote:
Maybe some warnings should be flaged as FATAL, and an exception would
be thrown at the end of the model processing ?
On Thu, Jul 3, 2008 at 8:12 AM, Mike Edwards
<[EMAIL PROTECTED]> wrote:
Kevin,
My understanding is that the initial model processing step of Tuscany will
only generate warnings for these types of error to enable the complete model
to get processed and so the maximum number of warnings can be generated in a
single pass.
What I'm not sure about is whether the runtime will then try to execute a
model that has generated a warning of this type. I hope that the runtime
would refuse to attempt to run the model, since it clearly has a severe
problem. I don't know whether the current runtime does refuse to run such
models, but my hope is that it would not do so.
Other views?
Yours, Mike.
Kevin Williams wrote:
I've noticed that loading a composite file containing duplicate
component names results in a warning. Wouldn't an exception make more
sense in this case? The Assembly Model spec does not allow
duplicates.
My 2c:
- the model processing should issue Warnings or Errors and continue
processing models after both
- the runtime should be able to start/stop/invoke an artifact with
Warnings, not an artifact with Errors
- we need to pay very close attention to what's a Warning and what's an
Error
- by artifact I meant composite/component/service/reference, for example
if my composite contains 4 components, 3 OK and 1 with an error, then do
we want to allow the 3 components that are OK to be started and invoked?
I'd say yes, but I understand that other people may say no :)
--
Jean-Sebastien
