From wupc@ccs.neu.edu  Tue Nov 19 23:43:47 2002
Delivered-To: lieber@amber.ccs.neu.edu
Date: Tue, 19 Nov 2002 23:43:47 -0500 (EST)
From: Pengcheng Wu <wupc@ccs.neu.edu>
To: lieber@ccs.neu.edu
Cc: "Skotthe@Ccs.Neu.Edu" <skotthe@ccs.neu.edu>,
	"Lieberherr@Ccs.Neu.Edu" <lieberherr@ccs.neu.edu>
Subject: Re: FW: I wouldn't be supportive of AOP?

I think that section 6 of Ivar Jacobson's paper covers a lot of
important issues that are being studied by AOP researchers right now.

If we replace 'service' in his paper with 'concern' in AOP (actually, I
view services to be functional concerns), we can have amazing maps
between his terminologies and AOP topics. For example, he wrote:
 "The following types of changes must be covered by our method:
  (i) a new service is incorporated, not requiring any change of the
existing services,
  (ii) an existing service or any of its objects must be changed, but
existing instances can follow the behaviour specified by the old class,
  (iii) as the second case, but the existing instances must be transformed
into new instances following their new classes."

The issue (i) is approximately implemented by the 'introduction' or 'open
class' facilities in AOP; (ii) is something like object-based AOP weaving
mechanism,i.e., objects can decide if they want to be advised at run time,
which I think is a very important issue for dynamic AOP
techniques; (iii) is like the regular AspectJ advice facility, in which
existing instances are automatically advised by some aspects if their
join points match some pointcuts/advice in aspects.

--Pengcheng