[aosd-discuss] dynamic AOP and dynamic languages

[Eric Bodden]

Hello.

> Ok, I agree that probably there have been a number of 
> attempts to implement AOP in a number of languages. However, 
> the question is why they never received their part of the 
> attention? Is there a sound reason why everything seems to 
> fall back to Java in the end?

I guess this has much more to do with the distribution of the base
languages and their use in research.

AOP has just made the step out of the research community into industrial
use and Java has been by far the language most researchers worked with
during the last few years. A reason for this again is probably that
since Java came out a lot of analysis and  transformation platforms have
been developed that enable easy language research and implementation on
a Java base. (I am not saying this is a good thing, but it's just the
way it is.)

Another fact is that quite some companies which use AOP today (I think
of IBM, JBoss, Glassbox, Terracotta, ...) use AOP in a Java setting,
mainly because they use it for web-based or at least server
applications. Those companies sometimes give money to research projects
building on Java (or Eclipse, ...). Microsoft has been showing
increasing interest in this field during the last one to two years and
indeed projects based on .NET have been spread since, again at least
partly due to funding.

So altogether it's much more a political/legacy issue than the question
of what would be the best approach. (Again, I am not saying that this is
a good thing.)

> Or would it be a good idea to try to publish my kind of 
> implementation to an AOP conference regardless?

I guess the best would be to decide yourself based on a careful
comparison to existing approaches. As previous posters wrote it's not at
all true that there are no AOP approaches based on dynamic languages.
There are actually quite a lot, cause as you noticed, AOP comes
realtively natural in those languages. That makes their implementations
also usually less intrusive and more lightweight. So my recommendation
would be to find those tools, do a careful investigation and then try to
identify your personal original contributions to the field. A paper just
stating "I have written [yet another] a dynamic AOP language" would have
no chance IMHO. (In particular that it often does make more sense to add
AOP features to an existing language with a rather large user base, if
it does not impact language design too negatively.)
 
> I guess what I'm trying to find out is whether there really 
> is a contribution to AOP in dynamic languages, and if there 
> is, whether it is worthwhile trying to convince the AOP community?

As I mentioned, that's the key point. I guess you are the only one who
can tell if there are any contributions and if so which.

Good luck with it.

Eric

--
Eric Bodden
Sable Research Group, McGill University
Montreal, Canada
http://bodden.de