Very Computer

Hi friends,

Just a thought: Would it be possible to enable linux to use 32bit win95
drivers.
Probably most hardware will support this platform --- and I think using
a 32bit driver should be more simple than a DOS one, just in case one
has some hardware that doesn't run under linux.

Yours
               Joachim

P.S: Please no flames --- I don't think win(whatever) is superior to linux !
--
Joachim Seifert                     |                
Tuerkenstrasse 28                   |              University of Munich
80333 Muenchen                      |              Theresienstrasse 37
Germany                             |              80333 Muenchen
----------------------------------------------------------------------------
If you think my .sig isn't funny, wait till you meet me !!!!!!!!!!!!!!!!!!!!!

The problem would be mapping the services Linux needs to the services the
driver provides, and mapping the services the driver needs of linux to
the services linux provides.  I'm not positive, but I think the two
OSs are sufficiently different for this not to work.  Certainly not
worth the effort.
Brian Hurt

: Hi friends,
:
: Just a thought: Would it be possible to enable linux to use 32bit win95
: drivers.
: Probably most hardware will support this platform --- and I think using
: a 32bit driver should be more simple than a DOS one, just in case one
: has some hardware that doesn't run under linux.
:
Problem with this idea. =) Win95 drivers are not ALL 32bit. =)  So we would
not gain anything. =) We would be  better off emulating BIOS drivers for
DOS under Linux.

Quote:>Problem with this idea. =) Win95 drivers are not ALL 32bit. =)  So we would
>not gain anything. =) We would be  better off emulating BIOS drivers for
>DOS under Linux.

I'll assume you know that a VxD is a 32-bit driver in ring 0.
They can call 16-bit code, but it is complex and time consuming.
Give us an example of how and why a typical VxD driver would go
to the trouble of calling 16-bit code and reduce performance,
when the whole point of VxDs is to improve performance over the BIOS?

Hi friends,

I don't want to appear as a windows advocate --- but if win95
drivers are *not* all 32bit, what about NT.
It's the future of win95 anyhow (after some $$$ for upgrading
for revolutionary features) --- certainly there are not many
drivers now, but there will be.
And it might be interesting for non x86-linux, as NT seems to
be present on many architectures.

Yours

                   Joachim

: : Hi friends,
: :
: : Just a thought: Would it be possible to enable linux to use 32bit win95
: : drivers.
: : Probably most hardware will support this platform --- and I think using
: : a 32bit driver should be more simple than a DOS one, just in case one
: : has some hardware that doesn't run under linux.
: :
: Problem with this idea. =) Win95 drivers are not ALL 32bit. =)  So we would
: not gain anything. =) We would be  better off emulating BIOS drivers for
: DOS under Linux.

that's right, but, since win95 and win-nt are merging together
(according to Bill's announcement, there will be one more
win95 release until the final merge), it would be quite desirable
to have something like ibcs2 for win-nt.
No, I'm not thinking about something like WINE emulating all NT-APIs.
Given the kernel threads of 1.3.29 and a specification of the NT-kernel
it should be possible to write an NT kernel emulator and to load on
top the various subsystems NT is consisting of.
Just my DM 0.02,
Norbert
_____

Joachim Seifert

NT drivers are NT drivers.  There is nothing nbit about them.  
Written i C (mostly) and compiled to the normal NT format
(``32 bit'')

Quote:>It's the future of win95 anyhow (after some $$$ for upgrading
>for revolutionary features) --- certainly there are not many
>drivers now, but there will be.
>And it might be interesting for non x86-linux, as NT seems to
>be present on many architectures.

Many Architectures?  I counted 4 (IA Alpha MIPS PowerPC) and
only very specific platforms.  There are more SGI/MIPS
platforms of Unix than NT ports.

Quote:

>Yours

>                   Joachim

Is such a beast really possible ---- sounds great !
Hope this will be a topic for discussion.

Yours

            Joachim

--
Joachim Seifert                     |                
Tuerkenstrasse 28                   |              University of Munich
80333 Muenchen                      |              Theresienstrasse 37
Germany                             |              80333 Muenchen
----------------------------------------------------------------------------
If you think my .sig isn't funny, wait till you meet me !!!!!!!!!!!!!!!!!!!!!

In newsgroup: comp.os.linux.development.system

Quote:

> that's right, but, since win95 and win-nt are merging together
> (according to Bill's announcement, there will be one more
> win95 release until the final merge),

You get points off for believing *anything* Bill says... that guy will
change his tune anytime he thinks there is a better way to*the
customer, and usually will deny ever having said anything in the first
place.

        /hpa
--

"The earth is but one country, and mankind its citizens."  --  Baha'u'llah

: >
: > that's right, but, since win95 and win-nt are merging together
: > (according to Bill's announcement, there will be one more
: > win95 release until the final merge),
: >
: You get points off for believing *anything* Bill says... that guy will
: change his tune anytime he thinks there is a better way to*the
: customer, and usually will deny ever having said anything in the first
: place.

that only shows he's a "true" marketing guy.
Anyway, in my opinion win95 is a genial invention to feed
application software to NT. Just have a look at actual
announcements. I'm sure in two years most win-SW will use
win32s and VxD-drivers.
But back to my original posting, is anyone out there beeing
able to comment on the possibility on emulating win-nt
kernel services in linux (comparable to ibcs2)?
Are these services documented or completely hidden by
layers as win32 or "tiny"posix?

ciao
_____

When it comes to sending data to the kernel, Windows NT has a good number
of differences from Unix.  While the basic concepts are very similar, the
big difference is the ability in NT to do multiple asynchronous writes
to a kernel driver and then later wait for an event to complete that tells
you that the request has completed.  The multiple asynchronous requests
are handled by an I/O stack in NT.

It would be alot more work to get NT drivers to work in Linux than the effort
get ibcs2 to work with other binaries.  Currently in Linux, there is no
support for spinlocks, mutexes, events, or a registry, all native NT objects
that a NT driver can make use of.  If Linux had all of these, it would still
be a difficult task because of the differences between NT and Unix I/O models.
Without these, it makes the task rather daunting.

Gordon Chaffee