From: John Max Skaller (skaller_at_[hidden])
Date: 2001-05-21 17:12:02
Gabriel Dos Reis wrote:
> | 'Global' policies are an evil idea, just
> However in our specific case, we're talking about functions operating
> on raw data, that is we're talking about datatypes.
> Any property shared by a group of data can be thought of as a "global"
> property (with appropriate meaning of "global"). And not just because
> a global data be conceived as "evil" in well-defined situations that
> it makes any property common to a group of objects an "evil idea".
I was being somewhat flippant of course :-)
I think of 'global/local' as the same abstraction.
In other words, anything shared by some things,
is also not shared with other things. I will speak of
the 'unit of localisation' as a way of nominating the
In the case of rounding, there are several
possible units of localisation:
2) per project
3) per program
4) per operation (statically)
5) per operation (dynamically)
I think the right choice here is (4). The next best
choice is (1).
I note that choices of units of localisation
characterize systems, and that people very often
ignore this. In particular, people regularly
use global data instead of parameterization,
and the choice is often wrong, but that is only
discovered later (when the requirements are extended).
-- John (Max) Skaller, mailto:skaller_at_[hidden] 10/1 Toxteth Rd Glebe NSW 2037 Australia voice: 61-2-9660-0850 checkout Vyper http://Vyper.sourceforge.net download Interscript http://Interscript.sourceforge.net
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk