[cap-talk] GC versus RAII

[Rob Meijer]

On Sun, May 1, 2011 10:25, Ben Kloosterman wrote:

>> Anyhow, OO 'purity' is a non-issue as far as I'm concerned. And GC for
>> values works very well in my model - doesn't interfere with resource
>> management, doesn't break the system's behavior model, and certainly
>> doesn't hurt developer productivity for managing things other than pure
>> values.
>
> Its not important to me either ,  breaking a few OO rules where
> appropriate eg for logging or resource management can simplify designs and
> make things easier for  programmers with lesser skills.

What you two are I think missing is that without purity of abstractions
(either OO or otherwise), complexity will leak trough your abstractions
and your abstractions are rendered inefective or often even
anti-effective.
I'm a big fan of multi-paragigm like for example combining OO with generic
programming, but only as long as the abstractions remain sufficiently
'pure' to remain usefull as abstraction. To me the 'inf memory'
abstraction that comes with GC should have been an ínf resources'
abstraction instead to be a usefull non anti-effective abstraction.