On Tue, Jan 18, 2000 at 04:02:36PM -0500, Andrew Brown wrote:
> >> actually...since i'm just mouthing off here, it would probably be
> >> "even better"(tm) to have a generic hw_input() routine that all the
> >> other *_input() routines (eg, ether_input(), ppp_input(), arc_input(),
> >> atm_input(), fddi_input(), ddp_input()...) could pass up to that would
> >> check something like this.  top-heavy, of course.
> >
> >you don't want this. We have a highly optimized ethernet compare function,
> >that does not need passing a parameter.
> 
> yes, but i was just making a (silly?) case for an abstraction layer
> that could provide a piece of (needed?) general functionality.  it
> seems to me that placing code in each ethernet driver to the compare
> would be silly, since the code would be indentical for each driver.
> and then to extend to the other hardware layers, since they might like
> to have the same option.
> 
> >Comparing one-byte ARCnet addresses is even easier, should it be needed.
> 
> sure.  but i've very little experience with other hardware layers, so
> i on't know in advance their characteristics (ie length, features,
> etc).
> 
> >We better keep this specialized.
> 
> ether_input() then?

Uhm... yes, I guess so.
	-is