At 18:35 Uhr +0200 10.06.1998, Olaf Seibert wrote:
>In <l03102805b1a340d73ed5@[192.168.1.4]> Hauke Fath
><hauke@Espresso.Rhein-Neckar.DE> writes:
>
>>Not including headers from other headers as a rule leaves at least a slight
>>chance of understanding for those unfamiliar with a program's structure.
>>Source files stay somewhat self-contained, and you can see what's imported
>>without following each and every header three levels deep.
>
>Including required headers from other headers as a rule leaves a lot
>more chance of understanding for those unfamiliar with the structure of
>internal depencencies of the header files. Header files stay completely
>self-contained, and you don't need to know which cruft to include just
>to appease the header file which contains the things you actually use.

Nice return.

But still, Open Source files are more often read than written. You can
always have a

/* Requires blah.h, sys/foo.h and mach/fuzz.h */

at the top of the header if you feel it's needed. Finally it all boils down
to having the necessary information about imports in the source file vs.
distributing them over several levels of headers.

	hauke


--
"It's never straight up and down"     (DEVO)