Re: [Boost-users] Question about Boost Interfaces

From: Jonathan Turkanis (turkanis_at_[hidden])
Date: 2008-02-01 11:55:55

Martin de Lasa wrote:
> Jonathan,

>     I never officially proposed the library; I just requested comments.
>     Although there seemed to be quite a bit of interest, I became very busy
>     and didn't have time to work on it.
> 
>     However, I'm working on it again now -- as of last week -- and hope to
>     request comments soon and then to propose it for inclusion.
> 
> 
> That's great. Out of curiosity are you shooting to have a working draft 
> ready for some particular time?

I'm thinking maybe April 1. Wait ... maybe I'd better make that April 2.

>     I don't know a lot about AOP; the support for AOP in the interfaces
>     library was based on another article by Chris and wasn't very well
>     developed. I think I want to include pretty comprehensive support, or
>     none at all, and I haven't decided which yet. Sorry I can't be more
>     helpful right now.
> 
> 
> It funny, though I have been annoyed for years by the design issues that 
> AOP attempts to address, I only became aware of AOP (as a formal 
> technique) recently. So, it looks like we are in the same boat !!
>
> Have you found any good papers/books/videos that you feel talk about AOP 
> at the depth of level for which you want to provide support? As with any 
> techniques I am sure the 80/20 rule applies to AOP (i.e. 20% of features 
> will deliver 80% of the benefit). Though I have never been part of the 
> formal boost lib proposal process, I wonder if getting a solid/minimal 
> subset of AOP features into a boost lib might do as a first cut to 
> proposing the lib.

Based on the AOP literature I've read, my feeling is that a direct 
"port" of an AOP framework from some other language that does not use a 
specialized pre-processor would not satisfy AOP people and would not fit 
C++ very well. My plan is to formulate a model that is designed 
specifically for the capabilities of C++ and that can provide solutions 
to (some of) the main problems that motivate AOP. I expect the notions 
of advice and pointcuts, for example, to be quite different from 
existing frameworks.

I have a bunch of ideas but I haven't worked them out yet to the point 
where I want to share them.

> Thanks again. I look forward to what you come up with!

Thanks!

> Martin

-- 
Jonathan Turkanis
CodeRage
http://www.coderage.com