[ On Wednesday, September 29, 1999 at 15:31:53 (+1000), Julian Assange wrote: ]
> Subject: Re: Bitfields and kernel
>
> Well, that's the end of struct ip.

Yeah!  :-)  Well, no, that would be silly, wouldn't it be.

> The structure of information must be defined somewhere. Without
> bitfields / structs you are defining this structure in a manner
> that spreads it ALL OVER the code that uses it and not in any
> central place. Consequently the structure is redunently defined.
> I shouldn't need to explain to this audience the evils of redundant
> code.

I can't agree more!  Indeed well thought out data structures that make
the best use of the idioms of the language at hand should always be used
where possible inside a program.

It's only that pesky external data that causes problems.  K&P in The
"Practice of Programming" chapter 9 outline several good techniques for
getting portable programs to communicate through portable external data.
(This kind of information was sadly missing even in a primitive form in
the first K&R, though obviously in a young language one has to be
somewhat careful about what one carves in stone! ;-)

Mapping pre-defined external data, such as chip registers into internal
data structures is sometimes best done with bit-fiddling because there's
no need to manipulate that data within the program.  Loading all of the
register bits into a bit-field and then storing them again after
changing only one or two of them would be an obvious waste of
programming, especially in a performance sensitive section of a device
driver.

Doing the right thing for something like struct ip is trickier because
experience has shown that its handling definitely needs to be optimised
at least to some degree.

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>      <robohack!woods>
Planix, Inc. <woods@planix.com>; Secrets of the Weird <woods@weird.com>