[TUHS] Bitsavers' RT/PC, AIX, AOS, etc. recent additions

Kevin Bowling kevin.bowling at kev009.com
Tue Feb 18 23:41:18 AEST 2020

Well, it’s not particularly hard to find a Mach system today.  The fruit
company makes computers and mobile phones with it.  Steve Jobs enamor for
Avie Tevanian is well chronicled in the common biographies.

The problem with Mach is their ukernel implementation was pretty shit while
the ideas were sound.  MacOS runs like an old dog to this day on anything
remotely system% intensive (like say a network server).

IBM abandoned the idea of any ukernel with AIX3 for RISC/6000.. Charlie may
be able to add commentary on that but it was almost certainly for
performance which was paramount in the workstation wars and RS6K had an
front runner opening.

The idea of small TCBs and IPC has really been perfected in L4 kernels;
seL4 in particular is one of the most interesting pieces of systems
software today.  Fruit company uses a different older L4 for their security
coprocessors and had a Darwin port for L4 at one point called Darbat.


On Tue, Feb 18, 2020 at 6:29 AM Jason Stevens <
jsteve at superglobalmegacorp.com> wrote:

> I was more interested in the "Mach" kernel itself as I've only recently
> been able to get it to boot up from sources for the i386.
> I hadn't looked into the other aos/vrm stuff.  But that is interesting, a
> 4.3 with the vfs.
> In hind sight maybe Mach wasn't so bad with its messaging and threads,
> along with multiprocessor support.. Its what we all were eventually
> desiring anyway.
> One thing is for sure, multiple GHz machines sure make it a lot easier to
> use, these days.
> I'd gotten lucky with Mach as the platform code is really modular and even
> a monkey like me banging on a keyboard of an existing Mach 386 machine was
> able to get the latter source running under the older platform code.  Shame
> Mach 3 seems to have broken all the fun stuff or requires real effort and
> understanding... Things I lack.
> But I was really surprised about the coprocessor cards..  I wonder what
> other interesting things are in there.  Or how hard it is to hammer 386 BSD
> into aos "sort of a 4.3 Tahoe ++"
> ------------------------------
> *From:* Kevin Bowling <kevin.bowling at kev009.com>
> *Sent:* Tuesday, February 18, 2020, 9:02 p.m.
> *To:* Jason Stevens
> *Cc:* Charles H Sauer; TUHS
> *Subject:* Re: [TUHS] Bitsavers' RT/PC, AIX, AOS, etc. recent additions
> Thanks for clarifying.  I will reassert that the three pieces of systems
> software I mentioned (VRM, AIX2, AOS) are not Mach in any way I know
> about.  AOS may have some generic cross pollination, it’d be whatever was
> going on at CSRG also for non-RT (4.2-4.3?) BSD platforms at the time of
> checkout. Kirk or Warner may be able to elucidate if provided the date and
> some reference material from AOS or I can do some original research.
> Most distinctly and important:  VRM is not in any way Mach, it was its own
> bespoke microkernel.  The microkernel would have been the most “Mach” part
> of Mach research, so this makes the VRM concept even more unique and
> enjoyable to me being so different and ambitious.  Therefore I don’t think
> it is particularly correct to say any of VRM, AIX, AOS software is Mach
> without its ukernel.
> What you linked is a very late port (late 1990s) of a hybrid of 4.3 and
> 4.4 BSD (late meaning in the time when Net, Free, and Open had long taken
> over from CSRG BSD).  I will quote a Twitter communication I had with Miod
> Vallat in the past:
> “Also it's not really 4.4. It's a mix of 4.3BSD-Reno plus the 4.4 VFS
> layer and new system calls. It still uses the 4.3, pre-Mach, VM system,
> hence no mmap(2).”
> What Miod means by “pre-Mach” above: 4.4 BSD adopted the kernel memory
> subsystem of Mach into the existing BSD monolithic kernel. Not any of the
> ukernel or things like Mach IPC.
> Not trying to be overly pedantic with you just trying to keep the records
> straight since these machines are one of my keen interests and I welcome
> new information on them.
> Regards,
> Kevin
> On Tue, Feb 18, 2020 at 5:30 AM Jason Stevens <
> jsteve at superglobalmegacorp.com> wrote:
>> Oh sure!
>> I'm having to use my phone...
>> It's the combined sources here:
>> http://bitsavers.trailing-edge.com/bits/IBM/RT/rt_bsd44/
>> doc  mk
>> jsteve at localhost:~/rt_bsd4/src/sys/.local/mach2.4$ pwd
>> /home/jsteve/rt_bsd4/src/sys/.local/mach2.4
>> jsteve at localhost:~/rt_bsd4/src/sys/.local/mach2.4/mk/conf$ cat vers*
>> 69
>> 5
>> 1
>> X
>> So 5.1x edit 69
>> jsteve at localhost:~/rt_bsd4/src/sys/.local/mach2.4/mk$ more CHANGELOG
>>  17-May-88 David Golub (dbg) at Carnegie-Mellon University
>>  XM21:
>>         David Black completely rewrote the accurate timing code
>>         (which is now implemented on all machines) and the priority
>>         and scheduling algorithms. The system now correctly reports
>>         cpu_usage per thread.
>> The all file has this before i386 was added.
>> So it's an older v2 than what is on the CSRG CD, but not as old as the
>> VAX '86 stuff.
>> It seems to be March 11 1989, although that could be when this was either
>> archived or ported..  I guess they didn't exactly sync to a public kernel
>> tree all that often.
>> On Tue, Feb 18, 2020 at 4:05 PM +0800, "Kevin Bowling" <
>> kevin.bowling at kev009.com> wrote:
>> I’m asking exactly where the Mach is in the linked archive. VRM, AIX or
>>> AOS? Can you support this with a reference for my own documentation
>>> On Tue, Feb 18, 2020 at 1:02 AM Jason Stevens <
>>> jsteve at superglobalmegacorp.com> wrote:
>>>> It's the CMU micro kernel.  The hybrid "2.6" lived on in NeXTSTEP, and
>>>> OPENSTEP, with various upgrades to bring it up to OS X.
>>>> The RT as I understand it was a research machine, hence the BSD ports,
>>>> and Mach port.
>>>> What is interesting the more I dig around is that there was ROMP
>>>> coprocoessor cards, and an OS/2 and DOS monitor program to let you boot BSD
>>>> on the card.  Peripheral IO was done on the x86 side.
>>>> If RT's are rare, I can't imagine how impossible it would be to get one
>>>> of those cards!
>>>> The BSD assembler and linker source is in the archives too, no doubt
>>>> it'll help someone make a RT emulator.
>>>> Get Outlook for Android <https://aka.ms/ghei36>
>>>> On Tue, Feb 18, 2020 at 12:54 PM +0800, "Kevin Bowling" <
>>>> kevin.bowling at kev009.com> wrote:
>>>> Can you clarify what is Mach in this archive if I have a gap in my
>>>>> knowledge? I didn’t know the VRM had any direct relationship to Mach
>>>>> Regards,
>>>>> Kevin
>>>>> On Mon, Feb 17, 2020 at 9:43 PM Jason Stevens <
>>>>> jsteve at superglobalmegacorp.com> wrote:
>>>>>> Interesting stuff!  And another version of Mach is buried in there.
>>>>>> So the 4 csrg cd set may have updates to the romp support as it's an
>>>>>> older version of the 5.1 kernel from 89...  Not that think there is any
>>>>>> Mach romp users.
>>>>>> Get Outlook for Android <https://aka.ms/ghei36>
>>>>>> ------------------------------
>>>>>> *From:* TUHS <tuhs-bounces at minnie.tuhs.org> on behalf of Charles H
>>>>>> Sauer <sauer at technologists.com>
>>>>>> *Sent:* Tuesday, February 18, 2020, 5:51 a.m.
>>>>>> *To:* TUHS
>>>>>> *Subject:* [TUHS] Bitsavers' RT/PC, AIX, AOS, etc. recent additions
>>>>>> The Bitsavers' RSS feed (
>>>>>> http://user.xmission.com/~legalize/vintage/bitsavers-bits.xml)
>>>>>> seemed to me to be dominated by RT, AIX, AOS (BSD for RT), etc. stuff in
>>>>>> the last week or so. I've only sampled a few items, but discovered a few
>>>>>> things that I should have known (or knew and forgot?) while I was at IBM.
>>>>>> http://www.bitsavers.org/pdf/ibm/pc/rt/ -- voice: +1.512.784.7526
>>>>>> e-mail: sauer at technologists.com fax: +1.512.346.5240 Web: https://technologists.com/sauer/
>>>>>> Facebook/Google/Skype/Twitter
>>>>>> <https://technologists.com/sauer/Facebook/Google/Skype/Twitter>:
>>>>>> CharlesHSauer
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://minnie.tuhs.org/pipermail/tuhs/attachments/20200218/5660f755/attachment.html>

More information about the TUHS mailing list