Johnny Billquist wrote:
Are you actually talking now about implementing a DECnet stack on a machine
in user mode?
That is definitely not doable.
Is that really true? I don't believe it. How about an Einstein style
"gedanken experiment" - put simh on a machine (which is an entirely user
mode program), load VMS on it. Bingo - a user mode implementation of a
DECnet stack.
I don't mean this as a practical solution to DECnet on Linux, but it's
clearly possible to get all the Ethernet access you need to implement DECnet
in a user mode program.
N.B. "user mode" is not the same as "run as root". Promiscuous mode
requires root privilege, but that's still a user mode program.
Bob
Just consulted a few online VPUs comparisons and it looks like we're both slightly wrong, although I'm happy to be corrected:
VAXstation 4000 model 60: 12 VUPS
VAXstation 4000 model 90: 32 VUPS
VAXstation 4000 model 90A: 40 VUPS
VAXstation 4000 model 96: 45 VUPS
However as I've said apart from CPU intensive tasks the '60 doesn't 'feel' much slower than the '90.
Regards, Mark.
Sent from my iPad
On 7 Jun 2012, at 14:12, hvlems at zonnet.nl wrote:
The Vaxstation 4000 model 60 is rated at approx 20 VUPS, the 90A at 40 VUPS. I don"t own a model 90. The 90A's are very fast compared to other Vaxes I worked with. Two exceptions, the 4705A and the 4108 (4105?, the one that cames in a plastic PC style cabinet) were also quite fast. The Vaxstation 4000 model 96 is faster than the 90A (45 VUPS). I certainly would like one for my collection!
Hans
------Origineel bericht------
Van: Mark Benson
Afzender: owner-hecnet at Update.UU.SE
Aan: hecnet at Update.UU.SE
Beantwoorden: hecnet at Update.UU.SE
Onderwerp: Re: [HECnet] Introducing myself... and my little network
Verzonden: 7 juni 2012 13:41
On 7 Jun 2012, at 10:33, Jordi Guillaumes i Pons
<jg at jordi.guillaumes.name> wrote:
Hello list,
I've just joined HECnet under the area 7, and I wanted to introduce myself and my little puppies :)
Hello!
I am interested to know how you find the performance of the VAXstation
4000/60 compared to others.
Good to see someone with such great skills on old iron join us. I know
very little outside vasics and simple DECnet setup :)
--
Mark Benson
http://markbenson.org/bloghttp://twitter.com/MDBenson
I told my wife many years ago that if I ever found an 11/750 or 11/780 it's coming to live in our garage. It was the first real computer I ever used. The largest VAX in my collection so far is only the size if a bar fridge so I think that's all she's expecting. I guess it's probably best that I haven't actually found an 11/780 yet...
Ian
Sent from my iPad
On 2012-06-07, at 5:35 AM, Jordi Guillaumes i Pons <jg at jordi.guillaumes.name> wrote:
Jordi Guillaumes i Pons
Barcelona - Catalunya - Europa
El 07/06/2012, a les 13:46, Dave McGuire <mcguire at neurotica.com> va escriure:
On 06/07/2012 07:26 AM, Jordi Guillaumes i Pons wrote:
According to this:
http://h71000.www7.hp.com/openvms/os/openvms-release-history.html
Support for the 4000/90 came in version 5.5-2 (1992). The 3300 was
supported in 5.0-2 and the 4000 mod. 2000 in 5.4-2. So none of my real
boxen can run 4.7.
Well you've gotta find yourself some more machines! 11/750s are
probably the most common of the 4.x-capable machines, though there don't
seem to be many left floating around.
Heh, then I would have to find also a new home and probably an attorney to handle de divorce demand ;)
-Dave
--
Dave McGuire, AK4HZ
New Kensington, PA
---
Filter service subscribers can train this email as spam or not-spam here: http://my.email-as.net/spamham/cgi-bin/learn.pl?messageid=ECEB3FF8B09C11E18…
On 7 Jun 2012, at 13:44, Jordi Guillaumes i Pons
<jg at jordi.guillaumes.name> wrote:
If I've understood it correctly, the idea is to replace the (dying) kernel support for DECNet under linux by userland code, so it the current module dies definitely due to kernel ABI changes we could still use decnet in the Linux boxes.
As Johnny pointed out, some root access is required for a daemon to
play with the ethernet but with that in place the rest would be
userland, yes. The key is no kernel module or kernel support is
necessary because it's going away.
I'd add as an additional benefit that moving decnet out of the kernel would make possible to port it to other unix flavours, like the BSDs and Mac OSX.
That was also part of my modus operandi. It would port to prettymuch
and OS that supports libpcap and MAC address spoofing (possibly
Windows?).
--
Mark Benson
http://markbenson.org/bloghttp://twitter.com/MDBenson
On 7 Jun 2012, at 13:32, Johnny Billquist <bqt at softjar.se> wrote:
Are you actually talking now about implementing a DECnet stack on a machine in user mode?
That is definitely not doable. You need atleast one part that talks raw ethernet, if you want access to the local ethernet.
No. What we are trying to work out is if (as it seems to be) it is
possible to implement a functional DECnet stack without having to
directly alterbthe kernel code (as is currently required) because
doing so limits portability and compatibility due to ever-shifting
sands of Linux kernel development.
And that part needs to run as root. No way around that.
We established that, yes. That's where my idea of using pcap and
running the interface in promicuous mode came out of.
Normal user programs should of course not need to be running as root, nor much around in the kernel, but there is nothing tricky about this. The DECnet stack is one process. Any other program talking DECnet needs to talk to the DECnet stack. That can be done in a number of ways. Unix sockets, TCP/IP sockets, shared memory, named pipes... Just pick something.
So with the daemon running as root it is possible to access it from
'userland' without root access. Excellent.
The ugly part is that you'll need to write all the software to handle all the protocols that sits on top of DECnet, such as a FAL listener, a FAL user client, CTERM listener and client, NICE, PHONE, MAIL... The list goes on...
You'll probably want to write a library with all the DECnet library primitives so that you have a nice API for the user level code to use. If you are lucky, the API in the current implementation can be adapted to your new implementation, which will make it possible to reuse a lot of code. I have not looked at how the current DECnet API looks like in Linux.
That seems like a sound idea to me. If the code already exists there's
no point on duplicating the effort unless it's required.
Anyway, this is a different story than any potential problems with promiscuous mode...
As I explained, my thinking was to use promiscuous mode to created a
DECnet stack that wasn't kernel dependant. That's where Dave and I got
hung up on potential small issues.
All clear there now, we can move on!
And you still need to be root to play with ethernet.
Indeed.
--
Mark Benson
http://markbenson.org/bloghttp://twitter.com/MDBenson
A vaxstation 2000 and a microvax 2000 run V4.7 too. With VWS instead of Motif at a reasonable speed.
Hans
-----Original Message-----
From: Dave McGuire <mcguire at neurotica.com>
Sender: owner-hecnet at Update.UU.SE
Date: Thu, 07 Jun 2012 07:46:10
To: <hecnet at Update.UU.SE>
Reply-To: hecnet at Update.UU.SESubject: Re: [HECnet] Introducing myself... and my little network
On 06/07/2012 07:26 AM, Jordi Guillaumes i Pons wrote:
According to this:
http://h71000.www7.hp.com/openvms/os/openvms-release-history.html
Support for the 4000/90 came in version 5.5-2 (1992). The 3300 was
supported in 5.0-2 and the 4000 mod. 2000 in 5.4-2. So none of my real
boxen can run 4.7.
Well you've gotta find yourself some more machines! 11/750s are
probably the most common of the 4.x-capable machines, though there don't
seem to be many left floating around.
-Dave
--
Dave McGuire, AK4HZ
New Kensington, PA
The Vaxstation 4000 model 60 is rated at approx 20 VUPS, the 90A at 40 VUPS. I don"t own a model 90. The 90A's are very fast compared to other Vaxes I worked with. Two exceptions, the 4705A and the 4108 (4105?, the one that cames in a plastic PC style cabinet) were also quite fast. The Vaxstation 4000 model 96 is faster than the 90A (45 VUPS). I certainly would like one for my collection!
Hans
------Origineel bericht------
Van: Mark Benson
Afzender: owner-hecnet at Update.UU.SE
Aan: hecnet at Update.UU.SE
Beantwoorden: hecnet at Update.UU.SE
Onderwerp: Re: [HECnet] Introducing myself... and my little network
Verzonden: 7 juni 2012 13:41
On 7 Jun 2012, at 10:33, Jordi Guillaumes i Pons
<jg at jordi.guillaumes.name> wrote:
Hello list,
I've just joined HECnet under the area 7, and I wanted to introduce myself and my little puppies :)
Hello!
I am interested to know how you find the performance of the VAXstation
4000/60 compared to others.
Good to see someone with such great skills on old iron join us. I know
very little outside vasics and simple DECnet setup :)
--
Mark Benson
http://markbenson.org/bloghttp://twitter.com/MDBenson
Jordi Guillaumes i Pons
Barcelona - Catalunya - Europa
El 07/06/2012, a les 14:23, Johnny Billquist <bqt at softjar.se> va escriure:
On 2012-06-07 12:42, Dave McGuire wrote:
On 06/07/2012 06:37 AM, Sampsa Laine wrote:
Howabout starting it as a daemon (running as root) via the boot
scripts, and have non-root users's programs access it via a
socket?
Cool idea - you could even have multiple hosts connect this way but
export one DECNET endpoint, i.e. run CTERM on Box A whilst FAL goes
to Box B, and MAIL to Box C :)
Yes, you could multiplex/demultiplex it any way you wanted. It would
open up all sorts of interesting configuration possibilities.
I think it would help if you actually explained a little more what you actually want to accomplish, and what you see as the problems...?
If I've understood it correctly, the idea is to replace the (dying) kernel support for DECNet under linux by userland code, so it the current module dies definitely due to kernel ABI changes we could still use decnet in the Linux boxes.
The process separation and distributed tasks are just a proposed improvement over the initial idea.
I'd add as an additional benefit that moving decnet out of the kernel would make possible to port it to other unix flavours, like the BSDs and Mac OSX.
Johnny
Jordi Guillaumes i Pons
Barcelona - Catalunya - Europa
El 07/06/2012, a les 13:46, Dave McGuire <mcguire at neurotica.com> va escriure:
On 06/07/2012 07:26 AM, Jordi Guillaumes i Pons wrote:
According to this:
http://h71000.www7.hp.com/openvms/os/openvms-release-history.html
Support for the 4000/90 came in version 5.5-2 (1992). The 3300 was
supported in 5.0-2 and the 4000 mod. 2000 in 5.4-2. So none of my real
boxen can run 4.7.
Well you've gotta find yourself some more machines! 11/750s are
probably the most common of the 4.x-capable machines, though there don't
seem to be many left floating around.
Heh, then I would have to find also a new home and probably an attorney to handle de divorce demand ;)
-Dave
--
Dave McGuire, AK4HZ
New Kensington, PA
On 2012-06-07 14:22, Johnny Billquist wrote:
On 2012-06-07 12:37, Sampsa Laine wrote:
On 7 Jun 2012, at 11:29, Dave McGuire wrote:
Howabout starting it as a daemon (running as root) via the boot
scripts, and have non-root users's programs access it via a socket?
Cool idea - you could even have multiple hosts connect this way but
export one DECNET endpoint, i.e. run CTERM on Box A whilst FAL goes to
Box B, and MAIL to Box C :)
Are you trying to suggest something like NAT for DECnet?
I'm not sure how easy that would be to do in DECnet, as it works in some
different ways than IP that might cause problems doing this.
Reading through things a little more...
Are you actually talking now about implementing a DECnet stack on a machine in user mode?
That is definitely not doable. You need atleast one part that talks raw ethernet, if you want access to the local ethernet. And that part needs to run as root. No way around that.
Normal user programs should of course not need to be running as root, nor much around in the kernel, but there is nothing tricky about this. The DECnet stack is one process. Any other program talking DECnet needs to talk to the DECnet stack. That can be done in a number of ways. Unix sockets, TCP/IP sockets, shared memory, named pipes... Just pick something.
The ugly part is that you'll need to write all the software to handle all the protocols that sits on top of DECnet, such as a FAL listener, a FAL user client, CTERM listener and client, NICE, PHONE, MAIL... The list goes on...
You'll probably want to write a library with all the DECnet library primitives so that you have a nice API for the user level code to use. If you are lucky, the API in the current implementation can be adapted to your new implementation, which will make it possible to reuse a lot of code. I have not looked at how the current DECnet API looks like in Linux.
Anyway, this is a different story than any potential problems with promiscuous mode...
And you still need to be root to play with ethernet.
Johnny
