Re: [abc] revision of CC submission

I've come down with some sort of flu, so unfortunately, I probably won't
be able to contribute much to the CC paper fixup, at least not for the
next few days.

I've thought a bit about ECOOP. The marks we got at CC were quite high,
so it's possible that the exact same paper would have been accepted at
ECOOP, with its more AOP-friendly program committee. The improvements
that Oege suggests can only improve the chances of acceptance, so my
opinion is that we have nothing to lose by improving it as much as
possible and sending it in.

Ondrej

On Sun, Dec 12, 2004 at 02:26:30PM +0000, Oege de Moor wrote:
> Having contemplated the reviews of the CC paper, I think we want to
> expand it by dumping the current sections 6 and 7, and instead have
> a substantial section making a comparison to ajc which shows why
> the architecture proposed here is a good thing. This will be needed
> for SP&E and ECOOP alike....
>
> 6. Comparison with ajc
> 6.1 Source characteristics
> (use SLOCcount for this)
> - How big are ajc and abc?
> - What are the relative sizes of the type checkers?
> - What are the relative sizes of the weavers?
> - How well is ajc separated from the Eclipse compiler?
> How well is abc separated from Polyglot?
> - Why does ajc use its own version of BCEL?
> Is Soot also specific to abc (answer: no)
>
> 6.2 Compile times
> - benchmarks:
> 3 pure Java (including abc itself)
> 5 aspect (LoD-sim, LoD-weka, nullcheck-sim, profiler, ants)
> a few synthetic benchmarks, where we can artificially vary
> the number of pieces of advice and the size of the base program
> - measure:
> times for ajc and abc just running semantic checks
> times for ajc, abc -O0, abc
>
> 6.3 Runtime of object code
> - what is the effect of Soot optims?
> compare ajc, abc, ajc+Soot
> - brief summary of further optims, ref to PLDI submission
>
> 6.4 Ease of code generation and weaving
> - choice of rewriting at source versus IR:
> * thisJP -> thisJPStaticPart trafo in Polyglot
> (cf how done in ajc?)
> - advantages of Jimple for weaving
> * lots of locals in LoD for ajc
> * can rely on nullcheck elim when doing
> lazy inits
> - other examples?
>
> 6.5 extensibility
> - brief summary and ref to AOSD paper
>
> 7. Other related work
>
> Implementation of AspectWerkz, JBoss, AspectC++.
>
>
>
>
>
Received on Mon Dec 13 13:47:11 2004