On Mon, Feb 04, 2002 at 10:19:36PM +0100, Oskar Sandberg wrote:
> On Mon, Feb 04, 2002 at 03:11:25PM -0500, Tavin Cole wrote:
> > I'm arguing for the simplification of the interface.
> 
> It's not really a simplification but a change - you want Buckets to be
> abstractions of fixed size buffers rather than files.

Ok :)

> > P.S. counting bytes isn't a big deal but if it's unnecessary, why do it?
> 
> Counting bytes adds a bit of code complication, but negligable runtime
> cost. So it makes more sense to implement it in the bucket (meaning it
> may be run redundantly) then to leave it to the user (meaning that the 
> implementation may redundantly duplicated).
> 
> I think it is worth having just to make the interface more flexible
> (IIRC the CLI client uses it for the progress indicator for example).

Well, I humbly disagree.  Support classes should accomplish a specific
purpose efficiently and do no more, when there is a reasonable system for
combining them to achieve desired results.
See freenet.support.io.CountedInputStream (there should be a
CountedOutputStream too).

-tc

_______________________________________________
Devl mailing list
Devl at freenetproject.org
http://lists.freenetproject.org/mailman/listinfo/devl

Reply via email to