Nicholas Lesiecki wrote:

On the subject of new team members. I think Chad would be an excellent
addition to the Cactus team.


Thanks for the recommendation, although I don't know how much time I will have either!

As for the proposal itself, I have 3 areas of feedback.

1) AW vs. AJ

I think I prefer the stability and maturity of AspectJ, but I understand
that it requires more tool support.

One question I have, how well do AspectJ and AspectWerkz play together?
Personally, I am planning on integrating AspectJ into our production system
in the near future. I would hate to be shut out of Cactus2 because of that
decision. (I'm not convinced that AspectWerkz is mature enough to use in
production.) I would actually cite AspectJ's tool support (crosscut
browsing, for instance) as a bonus for users committed to the idea of AOP.
The only users I would worry about would be those who did not wish to do AOP
except for Cactus.

We could consider a dual implementation approach (AspectWerkz and AspectJ)
as Chad Wooley did for EasyMock. This could be a lot of trouble, but might
be manageable if we delegate a lot of the actual code to POJO's.



I think Nick meant to type "VirtualMock" instead of "EasyMock" in the last paragraph. Supporting both could be a good idea. However, Vincent's current approach of directly using the AspectWerkz syntax (via xdoclet-style tags) would not be compatible with AspectJ. If the AOP syntax were hidden by the API, though, it would definitely be possible to support both. I still believe that this may be the best approach, but it would all depend on whether or not the API could be made flexible enough without exposing the AOP syntax.

I like the approach of pulling all possible logic out of aspects into POJOs, but this may be a moot point with the xdoclet-style syntax.

3) Obviously, bringing in Aspects opens up whole new realms of unit testing
possibilities. The question is, what will Cactus contribute on top of those?
Will we have our own mocking framework that uses AOP?



I think the best case would be if there were a flexible, independent framework for using mock objects which Cactus can use. VirtualMock might fill this role, but it is still in the early stages, and limited in some ways. I started a thread today on the jmock-dev mailing list to discuss a proposal for a standard, flexible mock-object API.

I'm looking foward to seeing cactus2 evolve, and would be glad to help.

Thanks,
Chad

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]



Reply via email to