----- Original Message -----
From: "Daniel F. Savarese" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Monday, October 22, 2001 6:32 PM
Subject: Re: New jar dependency for Cactus
>
> In message <01e701c15b16$c9eee7d0$[EMAIL PROTECTED]>,
"Vince
> nt Massol" writes:
> >I'd like to ask if anyone sees a problem for using AspectJ
> >(http://www.aspectj.org) with Cactus ? More specifically the license is
MPL
> >(http://aspectj.org/servlets/AJSite?channel=download&subChannel=license).
Is
> >that a problem ? I don't believe so, but just wanted to be sure.
>
> I don't think there's a problem. Redistributing the support libraries
> doesn't appear to conflict, but Jon, Sam, and others know better than I
> what licenses are compatible and which aren't. Is there a list of
> compatible licenses somewhere on www.apache.org (or did I just
> accidentally volunteer to put one together)?
>
> >If it works, I think a lot of jakarta projects may find aspectj very
useful.
>
> The main issue for me up until recently has been that AspectJ has been
> changing too fast to rely on for significant development purposes. Now
> that it's stabilized, my main concern is with efficiency (how good is
> the generated byte-code, how intrusive is the use of reflection, etc.).
> So I'd tend to use AOP with AspectJ for testing, where I can compile
> without aspects for production use. Using it in Cactus is probably an
> excellent fit. I just wonder how ready Java programmers are to adopt AOP
> at this stage. If you use AOP under the covers in a development library,
> users of the software aren't going to care. But if the development
library
> includes an aspect library requiring the use of the AspectJ compiler, I
> wonder if programmers will shy away from it because of the additional
> learning curve and its "non-standard" nature. At any rate, it will be
> interesting to see how AOP makes its way into Java, C++, and other OOP
> languages over time.
>
Indeed, I agree with all this. I'd like to try it on Cactus for :
- internal stuff: logging entries/exits of methods without having to write
log in each method,
- non-intrusively providing runtime check of Cactus setup : verify if
correct jars are on the classpath, verify properties files, ...
As you see, the initial goal is to try it non-critical things and see how
well it works and how easy to use. There is also a steep learning curve as
it is a different way of thinking. After that, as you have pointed out, I'd
like to use it to help write unit test cases (not sure exactly how yet ...
:) ).
One idea that I have for now is to use AOP to complement testing of
applications :
- we have unit testing with junit/cactus,
- we have functional testing with httpunit or others
- we are missing pattern testing ! And this could be done with aspectj. What
is pattern testing ? It will test the following kind of things :
- verify that any business method that is called as actually passed
through a controller (MVC model) first,
- verify that no JDBC code is used anywhere in the code except in such
class/method,
- verify threading,
- verify that log4j is used through a wrapper,
- ...
-Vincent
> daniel
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [EMAIL PROTECTED]
> For additional commands, e-mail: [EMAIL PROTECTED]
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
