4 Jun
2013
4 Jun
'13
6:57 a.m.
On 06/04/2013 12:43 AM, Ken Seefried wrote:
It not like ns32k or mc88k, where getting things
going means recreating
GCC support, figuring how to jump to ELF, etc. And yet, since 2009, no one
has picked up 80386, which pretty much tells you how really important it is
in the grand scheme of things.
One would reasonably think so, but it generally doesn't work out that way.
Embedded systems...say, the big graphical display and user interface on a
hospital ventilator/respirator (and I use this specific example because it's
a real-world system developed right up the road from me) aren't in a
situation where the developers run out and re-flash all of them whenever a
new Linux kernel is released. I'd also bet (but don't know for a fact) that
Boeing doesn't do that with their seat-back display systems in 767s either.
Most production environments...be it full-blown computers in a datacenter
or the Linux machine that runs a hospital respirator...don't just jump to a
new OS release as soon as it's announced. Some (many!) NEVER upgrade to a
new OS release. There is almost never a compelling reason to do so, that
would justify the time and expense to do so.
There is another factor. It turns out that most embedded projects that use
the 80386 are done in big companies, usually because they hire based on
resume' keyword searches, low stated salary requirements, etc. In other
words...people who aren't experienced enough to know why it's probably not a
good idea. Those guys either get promoted into middle management where they
can't do any more damage (and is probably what they wanted anyway), and then
someone else takes over. When that new person digs into the system, it's a
"big complicated unfamiliar mess", and that makes it even less likely that an
OS upgrade will happen simply because a new OS release has come available.
-Dave
--
Dave McGuire, AK4HZ
New Kensington, PA