#: by James Carman's words the mind was *winged* :#
I would have to say that I believe in the KISS (keep it simple, stupid)
principle. Proxies are just fine in many cases. The performance costs of
proxies vs. "weaving" are negligible compared to what the methods are
already doing, for the most part. Now, if you do run into a situation where
the proxies are just too darned slow for you, then you can tweak the system
by using a different approach AOP or runtime class generation (CGLIB or
Javassist). I would venture to say that you probably won't need to do it.
In other words, YAGNI. Don't get me wrong, I do like AOP for certain tasks,
but when a framework such as HiveMind gives you a simplified solution (like
MethodInterceptors), why not use it?
Sorry I have been understood this way. I was not advocating against Hivemind, Spring or some other
aop solutions. To me the initial post seemed to be a research for an intro in aop. From this point
of view I considered some clarifications were needed.
However, the KISS principle is here applied only to the framework development, as from my point of
view the users are required to in both situations to learn only the pointcut definition language (I
wouldn't bring into discussion a few interfaces API that are used in AW or AJ f.e.).
As regards the fact that proxy solutions are covering many cases I agree with you. But I usually
tend to use cleaner befores and afters, instead of always using arounds.
I don't want to start a flame, but considering interceptors as aop is not 100% correct. Interceptors
were introduced as a pattern, that at the point of aop concepts implementation seemed appropriate to
be used. You can do with them some of the things you are doing with aop, but that's all. Sorry if i
sound kinnda `religious´ about it :-).
:alex |.::the_mindstorm::.|
-----Original Message-----
From: Alexandru Popescu [mailto:[EMAIL PROTECTED]
Sent: Friday, July 29, 2005 5:35 PM
To: [email protected]
Subject: Re: Hivemind + AOP : Which approach to take?
#: by Vinicius Carvalho's words the mind was *winged* :#
In reply for James:
Haven't tried yet, promise I'll try over the weekend. And BTW Thanks a
lot for the post.
Alexandru:
"You can do offline weaving so there is no need for special
classloaders." My mistake, I'm so sorry for that, did not know about
that feature :D
I found some annoying problems with Annotations and eclipse plugin as well
:(
I have to agree that the plugin is not our powerfull point :-s. I would be
interested to hear what
problems have you faced with Annotations.
I think it's my fault the definition of intrusive on the email
context. As you can see by my name and grammar (English is not my
native language :P). What I'd like to say is: I prefer proxy based
once your original code is preserved, you don't need to re-compile it
to add/remove features from your aspects.
Unfortunately, the proxy based solution is not something I would recommend
to anyone. It is a very
easy approach (look how many so called aop solutions based on proxies are
out-there, compaired with
real aop solutions) and you are loosing some of the features (f.e. static
typing) which will finally
influence the performance. But this is quite another story.
Not justifying my mistakes, just making them clear :P
No problem. For myself it is quite the same: English is not my mother
language :-).
take care,
:alex |.::the_mindstorm::.|
Thanks all
---------------------------------------------------------------------
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]
---------------------------------------------------------------------
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]
