Re: Do we need this junk?

On Thursday 05 April 2007 11:39:41 am Nikolas Britton wrote:
> On 4/5/07, Peter Jeremy <peterjeremy_at_optushome.com.au> wrote:
> > [-stable removed since it's not relevant there]
> >
> > On 2007-Apr-05 04:58:17 -0500, Nikolas Britton <nikolas.britton_at_gmail.com> 
wrote:
> > >Can anything in the list below be removed from CURRENT?
> > >
> > >legacyfree1# cd dev/
> > >legacyfree1# grep -irsn isa ./ | grep -i include
> >
> > ...
> >
> > >legacyfree1# grep -irsn mca ./ | grep -i include
> >
> > ...
> >
> > Why do you believe anything in the list might need to be removed?
>
> I'd like to also add that 6-STABLE should be the last branch to support:
> 1. ISA / EISA
> 2. PC98 Platform.
> 3. i486
> 4. i586
>
> 98.83% of us have at least a i686 and 62.6% of us have at least a i786
> (SSE2) processor.
>
> Arch Break Down
> i386             5586     94.02%
> amd64 	        305       5.13%
> sparc64           30       0.50%
>
> x86 Break Down:
> i486	30           0.074%
> ??? 	51           0.125%
> i586	404         0.995%
> i686	14724	 36.230%
> i786	25431	 62.576%
> -----------------------------------
> Tot:	40640	100%
>
> data provided by bsdstats.org

Age alone is a lousy reason to drop support for any given piece of hardware. 
In fact, I consider the fact that it will install & run on whatever 
secondhand hardware you might happen to run across to be a major selling 
point of FreeBSD. As long as it's inclusion doesn't hamper advancements in 
other areas and there is someone to maintain it, support for more hardware 
new or old is always a good thing IMO. If you don't want to use it, take it 
out of your kernel config. The point of GENERIC is to cover as many different 
hardware setups as is reasonable, with emphasis on storage and network 
devices (without which it's difficult if not impossible to bootstrap or 
update a system). If you don't want to use the support then build a custom 
kernel without it. (But you don't lose much by leaving it alone.)

The numbers on bsdstats.org, while useful in demonstrating that there are _at 
least_ a given number of machines using certain hardware, should probably not 
be relied upon at this point for any other conclusions, especially regarding 
the (unknown but certainly a majority) portion of machines that are not 
represented.

In any case, patches speak louder than words. If you wanted to work on 
producing a highly functional legacy-free kernel tree (which maybe you are, 
for your unspecified new architecture mentioned in another thread), I'm sure 
that your work would not be ill-received.

JN