> >> one aimed at people who don't mind, nay, even enjoy, getting their
> >> hands dirty grubbing around inside kernels and compilers and such?
> 
> By "research system" I did not - do not - mean a system intended for
> people doing, say, psychophysics research, or aerodynamics research.
> Such people are, as you point out, point-and-drool end users from this
> point of view.
> 
> I meant a system intended for people doing software research.  Stuff
> like experimenting with tweaks to TCP to make it behave better in the
> current Internet.  Stuff like implementing IPv6.  Stuff like new
> optimization techniques in the compiler.

People doing software research _ARE_ point-and-drool users, when it
comes to the hardware that they're not hacking on, the software that
they're not hacking on, etc.

Why does some AI hacker give a flying fork() about anything other
than getting running and programming in a sane C user-land
environment?  Same for that compiler hacker.

Why does a networking researcher want to have to deal with vagaries of
installation, or with a short and quirky (or, at times, just plain
WRONG) "supported hardware" list?


Like i said before: researchers, including those doing low-level
operating system hacking, are really just "end-users," who do want
easy (perhaps even "point and drool") interfaces for installation,
administration, etc.  The reaserchers I talked to about this aren't
natural science researchers, etc.  I was talking to COMPUTER
SCIENTISTS, at conferences like USENIX and SOSP.

Indeed, NetBSD has lost a fair number of computer-science-research
users because of difficulty of the install process, difficulty with
"supported" hardware or standard hardware that is isn't supported,
etc.  There are people, on these mailing lists, who are NetBSD users
who have to work really hard to be using NetBSD, because the rest of
their groups have pretty much settled on other OSes for their
research, because of those factors.  That is NOT speculation, THOSE
ARE FACTS.


People doing software -- and hardware -- research have a finite amount
of time to blow on "overhead" of their OS: installation, maintenance,
etc.  They need to minimize that amount, so that they can get real
work done.  For that reason, until NetBSD is easy to install, supports
'standard hardware' well, etc., computer science researchers -- not
computer science hackers, computer science RESEARCHERS -- will and are
right to avoid NetBSD as their research platform unless they have a
local hacker to help them out with it.


cgd