RE: [abc] RE: Proof-reading OOPSLA paper

Hi Elnar,

Thanks for these very helpful comments!

I've implemented them as indicated below. Eagerly awaiting the next
instalment of your feedback...

>
> I have started reading the paper till Section 5. It looks pretty good so
> far, very comprehensible, great examples, impressive numbers. I especially
> like the offensive comments on PQL... :) Below are few minor suggestions.
> I'll read the rest of the paper tomorrow morning and email my further
> suggestions ASAP.

thanks

>
> you say abc-an extensible compiler for aspectj both in abstract and
> introduction. seems a bit redundant to me.

I think that's ok.

> page 2: we collect and develop of a set of...: I guess the first 'of' is a
> bit unfortunate there...

corrected

> page 4: the way the patterns are specified in J-LO: I think 'the' should
> not
> really be there

corrected

>
> page 4: AspectJ only allows one to intercept one event at a time...: the
> first 'one' seem to be redundant here

corrected

>
> page 4: you say: a simple aspect might keep a couple of identity hash
> maps:
> and then it turns out that it is actually three. Probably using the word
> several is better than a couple

corrected

>
> page 4: It also possible, however, to use aspects to do exactly what a
> good
> programmer would do by hand: 'is' is missing at the beginning of this
> sentence

corrected

>
> page 4: As said, it is useful to have this type of non-trivial
> implementation as a 'gold standard' in our benchmarks: I think it is
> better
> to start this sentence with 'As we said'

corrected

>
> sometimes you write 'section X' with a capital letter and sometimes not

all such references (similarly for figures and tables) should be capital
 
> page 4: Not all systems provide for free variables to be bound in the
> matching process: i suggest to rephrase this sentence

changed to:

In some systems, one can bind free variables via the pattern matching
process. When available, this feature greatly simplifies writing patterns
that track the behaviour of individual object instances. It is very hard to
implement such variable binding efficiently, however.

> page 5: ...but it is also hard to implement efficiently.: does not sound
> well

see new para above

> page 5: ...which indicated that its performance ranks well below those of
> the systems we benchmark here...I think it is better to remove the word
> 'well' because at the first look it gives a rather positive feeling about
> HAWK

well := considerably

>
> Page 6: The performance number listed was obtained by using a heap size of
> 1.5G: Is it just one number or numbers? what is 1.5G?

clarified: "listed for {\sc AjNaive} was in fact obtained"

>
> page 7: These highlight the importance of...: the importance of what?
> challenge N5?

clarified: "These highlight the importance of the next challenge: "

>
> page 9: Binding threads is impossible in PQL at the current time: I think
> it
> would be better just to say: Binding threads is currently impossible in
> PQL.

corrected

>
> page 9: strat-egy: is it a correct division of the word?

erm, just living with TeX's word breaking. I don't understand the rules!
>
> page 9: .and it cannot be expressed in PQL for the same reasons as that
> example: for the same reasons as IN that example

corrected

>
> page 9: i suggest to use * for the cases where the example is expressible,
> but has bugs and - for the cases where the example is not expressible. it
> seems more logical to me

I disagree: "*" is unfortunate but honourable, "-" is shameful ;-)

>
> page 9: On the NullTrack benchmark, using an appropriate data structure
> for
> organising the set of partial matches yields to a speedup of a factor of
> 6.:
> it is not clear what yields what?

deleted erroneous "to"
 
> Section 4: for example, Weka is written using three different fonts. is it
> justified? Same applies to APPROVE

all should be \sc, corrected in this section.

> > -----Original Message-----
> > From: Pavel Avgustinov
> > [mailto:pavel.avgustinov@magdalen.oxford.ac.uk]
> > Sent: 16 March 2006 20:10
> > To: aske@brics.dk; elnar.hajiyev@comlab.ox.ac.uk; dsereni@gmail.com
> > Cc: abc@comlab.ox.ac.uk
> > Subject: Proof-reading OOPSLA paper
> >
> > Hi,
> >
> > As you no doubt know, we have been working on an oopsla
> > submission about trace monitoring features. The deadline is
> > tomorrow, and we think we finally have a coherent enough
> > draft to request 'outside eyes' to have a look.
> >
> > I have put a current (at the time of writing) version at
> > http://musketeer.comlab.ox.ac.uk/~pavel/paper.pdf -- it is
> > almost complete, but still misses a conclusions section, and
> > some of the related/future work is not expanded.
> >
> > It would be great if you had the time to read through and
> > point out any typos, inconsistencies, unnecessary repetitions
> > or underexplained concepts that you can see. Of course, any
> > sort of comment is helpful -- preferably in an email to abc@comlab.
> >
> > As I said, the deadline is very close, so we need any
> > comments as soon as possible, tonight or tomorrow morning at
> > the latest.
> >
> > To avoid biasing your judgement, I shall say no more. :-)
> >
> > Cheers,
> > - P
> >
Received on Fri Mar 17 08:48:13 2006