>However, its not even close to being feasible. Three things that I
can't=20
>see anyway around:

>         1) VMS's shared library mapping technique wouldn't
>            work under NetBSD.

Why not? During image activation you have to map in
all the shareable images and then when each call is hit
for the first time, trap and fix it up. (At least, that's
my recollection fo what VMS does). In principle, it looks
feasible - a few man-years of work maybe. Is there something
fundamental that NetBSD could not do here?

>         2) The trap handlers are all different (VMS's appear
>            to be based on RSX like interrupts with fixed pools)

Do you mean ASTs here or condition handling/signalling? Assuming
this is VMS images on corresponding hardware (i.e. OpenVMS VAX on
NetBSD/VAX) then at least the hardware side is all there and you
just need to implement the software :-) ASTs, at first glance (and
with scant knowledge of netbsd internals) just look hard. The signalling
stuff requires trapping conditions (which must happen anyway in some
form)
and then walking a stack. I guess it boils down to whether you can=20
squeeze am OpenVMS-compliant stack into a NetBSD process.

>         3) The I/O system is based on a multi-threaded asynchronous
>            I/O model (the QIO) which could not be emulated in the
>            current I/O layer.

If NetBSD cannot do multiple asynchronous I/O then you are stuffed.

Even if it can, the real problems will probably lie in the fact that
many OpenVMS images rely on "expected" OpenVMS behaviour: logicals,
DCL behaviour, device driver behaviour and so on. Pretty much the same
sort of problems you would have trying to run a Unix image
on OpenVMS ("wot - no pipes?", "I'll just fork a few hundred times
...").

>I will be interesting to see if the VMS/Itanic project ever sees the
light=20
>of day (actually it will be interesting to see if the Itanic becomes
more=20
>than a marketing curiosity :-)

I doubt that Intel & HP can let Itanic fail - they have sunk too
much into it. The OpenVMS team are a pretty bright bunch - they'd
have seen the major showstoppers coming ages ago. If it doesn't
ever see the light of day that will be down to the bean counters,
not the engineers, IMHO.

Antonio