From: Gennadiy Rozental (gennadiy.rozental_at_[hidden])
Date: 2005-04-26 09:40:43
>>>I'm not saying that flexibility or genericity is bad. But it comes
>>>with a price.
>> What kind of price? Perforamce, usability, space?
> End-users work in a problem domain with its own vocabulary and semantics.
> Libraries / frameworks such as boost have their own, and so there's
> a potential for an impedance mismatch.
> The more a user is required to learn about aspects unrelated to his own
> problem domain just because the tools he wants to use are generic and
> to a broad range of domains the larger is this gap. It's a trade-off.
Yes. To use a tool you may need first invest time to learn it. And yes it's
a trade-off. But:
1. In case of plain memory management PBSP present exactly the same semantic
as any existent solution. So we could expect the "ordinary" user to be
familiar with it and do not require time to learn how to use it.
2. Above also imply that it shouldn't be that difficult to make one more
step to learn about possibilities of configuring it's behavior.
3. Resource management tasks are quite wide-spread. So whatever time you
spend getting hold of PBSP the time saving of writing specific simple
policies will overcome it. So the trade-off is quite reasonable IMO.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk