Cristiano Guadagnino 3.9.2005
Hi James, Samuel.
2James: You can install OS/2 or eCS from a bootable CD, at least with a fairly recent release.
2Samuel: It's great to have a coder with kernel programming experience!! L4 is a particularly
intriguing kernel. Our main worry is about device drivers... we thought about basing on ReactOS
kernel because that would allow us to use the immense availability of Win Device Drivers. But I
(we) would like to know more about your ideas (I especially like what you're saying about running
DDs as normal processes... but what about performance?). Note BTW that we want to maintain
as closely as possible all the features and the "feel" of OS/2. Binary compatibility would be best.
Yuri? Want to add/correct something?
Old Forum messages
Re: Old Forum messages
Samuel A. Falvo II 3.10.2005
Haha -- I totally didn't expect anyone to respond so quickly, and to me directly. I'm currently
writing an implementation of the Forth programming language and operating system using L4
as my foundation, so I'll be able to demonstrate code and performance at a later time. But
from what my research has shown, and if you implement your interfaces with care, you can
be quite performance competitive with traditional monolithic kernel designs using L4. L4's
messages can involve register values, copied strings from memory, memory mappings (which
can be revoked at any time later on), *or any combination of the above*. Also, the kernel
lumps as many useful operations per single microkernel call as possible. Another area that
interests me (even compells me) is "object capability" security systems. ACLs are nice, but
object capabilities are where you'll find true computer security. L4 provides the mechanisms
to make these work. Whether they can be used in a traditional OS environment like OS/2
remains to be seen, however. But I fully intend to explore them in my Forth environment.
Haha -- I totally didn't expect anyone to respond so quickly, and to me directly. I'm currently
writing an implementation of the Forth programming language and operating system using L4
as my foundation, so I'll be able to demonstrate code and performance at a later time. But
from what my research has shown, and if you implement your interfaces with care, you can
be quite performance competitive with traditional monolithic kernel designs using L4. L4's
messages can involve register values, copied strings from memory, memory mappings (which
can be revoked at any time later on), *or any combination of the above*. Also, the kernel
lumps as many useful operations per single microkernel call as possible. Another area that
interests me (even compells me) is "object capability" security systems. ACLs are nice, but
object capabilities are where you'll find true computer security. L4 provides the mechanisms
to make these work. Whether they can be used in a traditional OS environment like OS/2
remains to be seen, however. But I fully intend to explore them in my Forth environment.
Re: Old Forum messages
Samuel A. Falvo II 3.10.2005
An idea -- it seems to me that IBM took a logical set of steps to arrive at WPS-level
functionality. The first step was clearly the console-only environment. That is to say,
PC-DOS 4.0 was later renamed OS/2 1.0. If this is what you folks are striving for (e.g.,
starting with OS/2 1.0 functionality first as a milestone and working your way up from
there in chronological order), then I should tell you that I'm actually working on a
kinda sorta 32-bit BIOS-like environment that my Forth environment will run on top of.
Although my Forth runs co-resident with this library of functionality, I can write (once
the Forth code is done) an L4 server to serve as a BIOS for your own code environments.
And then there is the distinct possibility of adjusting the Forth system to be ANSI-compliant
and supporting OpenFirmware PCI cards! Hmm...
An idea -- it seems to me that IBM took a logical set of steps to arrive at WPS-level
functionality. The first step was clearly the console-only environment. That is to say,
PC-DOS 4.0 was later renamed OS/2 1.0. If this is what you folks are striving for (e.g.,
starting with OS/2 1.0 functionality first as a milestone and working your way up from
there in chronological order), then I should tell you that I'm actually working on a
kinda sorta 32-bit BIOS-like environment that my Forth environment will run on top of.
Although my Forth runs co-resident with this library of functionality, I can write (once
the Forth code is done) an L4 server to serve as a BIOS for your own code environments.
And then there is the distinct possibility of adjusting the Forth system to be ANSI-compliant
and supporting OpenFirmware PCI cards! Hmm...
Re: Old Forum messages
Chris Walton 3.10.2005
Personally, I'd like to see L4 as a backend instead of ReactOS for exactly those reasons that
Samuel mentioned. For Windows drivers, the ReactOS people are probably very open to help
porting their driver subsystem to the L4 Kernel like described above, in exchange for help in
porting the OS/2 stuff to ReactOS. One hand washes the other.
Personally, I'd like to see L4 as a backend instead of ReactOS for exactly those reasons that
Samuel mentioned. For Windows drivers, the ReactOS people are probably very open to help
porting their driver subsystem to the L4 Kernel like described above, in exchange for help in
porting the OS/2 stuff to ReactOS. One hand washes the other.
Re: Old Forum messages
James McCarte 3.14.2005
Trying to get OS/2 installed on a Dell Dimension system with no floppy drives, and am not
having any success. I can make the bootable CD, but when I tell it to boot from the installation
CD, it shows a message that says "Loading serial.?" (can't remember the extension), and hangs.
Any ideas?
Trying to get OS/2 installed on a Dell Dimension system with no floppy drives, and am not
having any success. I can make the bootable CD, but when I tell it to boot from the installation
CD, it shows a message that says "Loading serial.?" (can't remember the extension), and hangs.
Any ideas?
Re: Old Forum messages
Cristiano Guadagnino 3.15.2005
Hi James. This is not the right place to ask. I'd suggest using the forums on www.os2world.com.
BTW, you should provide more details: which versiorn of OS/2 are you trying to install? Bye
Hi James. This is not the right place to ask. I'd suggest using the forums on www.os2world.com.
BTW, you should provide more details: which versiorn of OS/2 are you trying to install? Bye
Re: Old Forum messages
Alex Bervoets 3.15.2005
IBM's microkernel (OS/2 for PPC) was based on the MACH kernel, so why shouldn't we do the
same ? Another suggestion : maybe we can convince IBM to release the source code of 'Workplace
Shell for Windows' under an open source license. Actually, WPS for Windows is freeware written
by IBM employees.
IBM's microkernel (OS/2 for PPC) was based on the MACH kernel, so why shouldn't we do the
same ? Another suggestion : maybe we can convince IBM to release the source code of 'Workplace
Shell for Windows' under an open source license. Actually, WPS for Windows is freeware written
by IBM employees.
Re: Old Forum messages
Yuri Prokushev 3.17.2005
2Cristiano: Personally I want to have most of OS/2 features recreated. Startiing from IFS level and higher.
Want to have stdio, vio, gdi, pm, som/wps, rexx, mmos/2, etc. Most of components created just excelent.
Some not finished but general idea is good. Binary compatible is also preffered. Only hard thing for
bincompatability is 16-bit calls. But such thing can be solved by virtualization of 16-bit code (not looked
closely yet, just idea). Actually, ReactOS is not good kernel anymore because, as mentioned by Sergey,
memory managment issues.
2Cristiano: Personally I want to have most of OS/2 features recreated. Startiing from IFS level and higher.
Want to have stdio, vio, gdi, pm, som/wps, rexx, mmos/2, etc. Most of components created just excelent.
Some not finished but general idea is good. Binary compatible is also preffered. Only hard thing for
bincompatability is 16-bit calls. But such thing can be solved by virtualization of 16-bit code (not looked
closely yet, just idea). Actually, ReactOS is not good kernel anymore because, as mentioned by Sergey,
memory managment issues.
Re: Old Forum messages
Samuel A. Falvo II 3.17.2005
Mach is a decent microkernel, but is slower than L4 by a rather significant amount (based on
published benchmarks). The L4 version I'm using is L4Ka::Pistachio, which supports the next
generation (and processor agnostic) API. It also supports a number of processors out of the box,
and supports SMP out of the box as well. Nonetheless, it is still quite fast, as the documented
benchmarks show. I am not intimately familiar with the Mach series, unfortunately; but I am
patently aware that Mach has, at least in the past, single-handedly given microkernels such a
bad name that few OS developers today consider basing anything on them.
Mach is a decent microkernel, but is slower than L4 by a rather significant amount (based on
published benchmarks). The L4 version I'm using is L4Ka::Pistachio, which supports the next
generation (and processor agnostic) API. It also supports a number of processors out of the box,
and supports SMP out of the box as well. Nonetheless, it is still quite fast, as the documented
benchmarks show. I am not intimately familiar with the Mach series, unfortunately; but I am
patently aware that Mach has, at least in the past, single-handedly given microkernels such a
bad name that few OS developers today consider basing anything on them.
Re: Old Forum messages
Samuel A. Falvo II 3.17.2005
Oops, I forgot to also mention that L4 also supports the necessary operations required for things like
orthogonal persistance, which would be a marked improvement for some classes of applications over
the traditional process model employed by most OSes. This is one of the best features of the IBM AS/400
series, for example. Of course, should OP be implemented in OS/3, it would definitely be a very long-term
goal into the future. To date, I am unaware of Mach offers anything even remotely close to this feature set.
Oops, I forgot to also mention that L4 also supports the necessary operations required for things like
orthogonal persistance, which would be a marked improvement for some classes of applications over
the traditional process model employed by most OSes. This is one of the best features of the IBM AS/400
series, for example. Of course, should OP be implemented in OS/3, it would definitely be a very long-term
goal into the future. To date, I am unaware of Mach offers anything even remotely close to this feature set.