[aosd-discuss] Non academic ideas on AOSD

[Radek Jedrasiak]

Hi Denis,
>
> > .... real world systems
> > can heavily suffer from some uncontrolled changed code.
>
> I agree with your points. And I do fear to apply AOP (as it exist today) in serious projects because it's
> hard to control the changes injected by advises.
>

Ufff ... i'm not the only one. Still i'd wish AspectJ 4.0 was already here <g>

>
> Consider a large project in development and some originally unanticipated crosscutting change that is to be
> made now. AspectJ approach would be to write an advise that "makes" the change. But several problems arise
> here, and the two of them bring the most danger:
> 1) the definition of the pointcut is tricky, since it must
> carefully select from a variety of potential join points, which can be numerous, hidden, tangled, and bearing
> unique implicit contracts;

That's a good one. You are right ! Why on earth would developers want to deal with
the syntax tree of a language or the hierarchy tree of ALL classes/constructs they
are probably about to "attack" by means of an aspect!?
This is something I did not think about until now. When developing software,
Aspects (in the form they work today AspectJ f.e.) add the need for knowledge
about the tool i'm using, which is not necessery until now: the explicit knowledge
of the language syntax to define joint points. Hmmm...

Cheerio
Radek