Could someone make brutally clear to me why it is that MD fixes to some of the
worst problems of BSD are not being allowed into the source tree?

I once asked a fairly important contributer to the project whether I would
be able to contribute code I had written for various devices.  I was told that 
if I had written decent quality code, it would probably be included.  

Now the type of code -I- write is still unworthy of kernel structures, but I've
taken a look at some of the machine dependent code that's been written, among
other things, to implement bounce buffers for the i386 port, and I can't really
understand why it's not being implemented, now that it's becoming so important
in an era where 16M of memory costs less than a hundred dollars, but a new
motherboard and CPU or new SCSI card costs easily twice that, and a large number
of the users of this project aren't making $30k yet.

Would someone please explain to me whether everyone shares the attitude
that if it's not MI, it should not be implemented whatsoever? 

I've heard a lot of things said back and forth between OpenBSD, NetBSD, and
FreeBSD.  I've been very happy with NetBSD, excepting the political arguments 
that seem to spread across mailing lists and newsgroups.

If bounce buffers are never going to be implemented in an MI manner, 
if we're going to have to maintain an army of patchfiles on a hundred different
ftp servers, please let me know.  

Because I installed RedHat linux on a machine this week, and I see a lot of features
I'd like to implement on NetBSD.  But if the niceties that literally every 
other PC unix have implemented are not ever going to be implemented by NetBSD,
I really can't see why I should bother.  I joined the NetBSD project two years
ago because the system was orders of magnitude better than what I was using at 
the time, and because of a small project to allow raytracing across differnt
platforms on differenct machines anywhere on a network.

I'm still going to put NetBSD on my production machines, where I have to have
security, where I need speed, where I need reliability, most importantly, where
I can specify hardware.  I'm not going to put SCSI on them, because I won't
be able to afford new cards probably, so I'll just put decent IDE.  

But if these problems cannot be worked out, when I'm working at home, the 
machine I'll be hacking on will be running FreeBSD or Linux or OpenBSD, because
I do not want to have to sup down a tree, and reimplement patches just so the
system I have, which hasn't changed hardware wise in weeks, will continue
to work with the newer kernel.  

Perhaps I'm way out of line.  If I am, tell me why.  But don't tell me that
the things that are turning away potential i386 users couldn't be fixed for
want of politics and personal preferences.

spb