[HECnet] Multinet alternatives ...

Wed Feb 26 12:04:24 PST 2020

You're quite right, Tops-20 directory creation /is/ a pain in the neck 
to the modern user.

However, a Tops-20 directory is a _very_ different beast than what you 
would see in other operating systems.  It was a different time with 
different costs.  Under DOS, Unix (all flavors), OS/2, Windows and IBM 
z/OS (nee MVS) a directory is simply a place to put files.  So you say 
make me a directory and /poof/, you're a directory.

Under Tops-20, directories have far richer semantics.  First of all, 
they are the store of record for login credentials, which is quite 
different from all the above and Tops-10, as well.  If you create a 
directory--/anywhere/--and it is not FILES-ONLY and it has a password, 
then you have just created a user. Furthermore connecting to a directory 
is in no way like doing a CD.  Once you are connected, you have the same 
rights as the owner and can use these rights to access files in other 
directories that the owner has rights to.  Group management is superior 
to Unix in that a directory can both be in multiple groups and provide 
multiple groups.  It allows for far greater granularity.  Disk usage is 
carefully kept track of.  However, management of all of it requires 
special administrative capabilities to be given.

All of this was necessary in the days of machines have 10's of thousands 
of users with strapped storage; I believe the default at Columbia for 
undergraduates was 100 pages (about 350 KB) permanent storage and 1,000 
working.  You've got to do things like that when all you can squeeze 
into the machine room is a couple of gigabytes worth of disk drives.  
The careful group management was necessary to keep the more inquisitive 
students from interfering with others' activities.  We needed special 
programs to manage all the accounts, the groups and the storage.  Most 
sites had home grown software like this and Columbia's was pretty nifty.

The only thing lacking was the ability for a user to create their own 
sub-directory.  I began looking at this by sketching out an IPCF% based 
client server paradigm.  The user would run a client to create the 
directory and then the server would do all the automatic nice things.  
It grew to be quite complex as it had to go recursive to grab storage 
from superior directories' allocation and then it would have had to 
interface into the group management system to request groups for the 
user to access the subdirectories.  I gave up when the line was 
cancelled and we went to Sun SPARC.

Non-DEC versions of Tops-20 (such as Stanford and PANDA) implement 
something called super-domestic structures where the user can always 
access their subdirectories.  A limited version of relative directory 
naming (viz, ./., ../) was implemented, also.  Columbia's group 
management software was intelligent and flexible enough that we never 
needed to use the super-domestic code.

Structure (and tape) mounts are also a very different thing, but that's 
another story.

Maybe I'll look at it again after I finish my DAP changes and the new 
FTP server.

------------------------------------------------------------------------
On 2/26/20 8:15 AM, Keith Halewood wrote:
>
> I managed to get the KLH10 emulator compiled and running the Panda 
> distribution of TOPS-20 under Raspbian on a Pi 4 without too much 
> trouble. Its ‘dpni20’ process has no trouble talking to a TAP device 
> which is bridged with eth0 and thence onto the mixed 
> in-memory/real-ethernet DECnet connection to the HECnet. Admittedly, 
> I’m not that familiar with TOPS-20 or the ‘hardware’ on which it runs. 
> My main interest was MDL at the time. It’s still running as WALACH on 
> the HECnet. My main ‘take-away’ from the experience was: “HOW many 
> questions do I have to answer in order to create a directory on this 
> thing??”
>
> On the SIMH side of things, I’m a bit stuck getting TOPS-10 running 
> from the DO scripts from the URL…. [I appear to have misplaced this 
> bit of information]
>
> Keith
>
> *From:*owner-hecnet at Update.UU.SE [mailto:owner-hecnet at Update.UU.SE] 
> *On Behalf Of *Rob Jarratt
> *Sent:* 26 February 2020 11:51
> *To:* hecnet at Update.UU.SE
> *Subject:* RE: [HECnet] Multinet alternatives ...
>
> I can’t remember now what the problems were, but I think I had 
> problems getting Phase III running on SIMH with the KMC or KDP (or 
> whatever it was).
>
> *From:*owner-hecnet at Update.UU.SE <mailto:owner-hecnet at Update.UU.SE> 
> <owner-hecnet at Update.UU.SE <mailto:owner-hecnet at Update.UU.SE>> *On 
> Behalf Of *Thomas DeBellis
> *Sent:* 25 February 2020 20:14
> *To:* hecnet at Update.UU.SE <mailto:hecnet at Update.UU.SE>
> *Subject:* Re: [HECnet] Multinet alternatives ...
>
> What issues had you encountered talking to Tops-20??
>
> If you are using the SIMH emulator for a KS processor, then the latest 
> version of Tops-20 you can run is 4.2, which I believe will do Phase 
> III.  Since there was never any Ethernet adapter for the KS (it was a 
> massbus device), I believe you must speak DDCMP, which we used with 
> KMC's, which would do 56Kbs.  There was a slower adaptor for 9.6Kbs, 
> but the name escapes me at the moment.
>
> I've had little trouble with Tops-20 V7 DECnet; that is, once I got it 
> to working.  When I was modifying FAL/DAP to accept anonymous 
> transfers, I and and others beat on it quite a lot to wring the bugs 
> out.  These regression tests were against another 20, RSX 11M+ (or 
> whatever MIM:: is running) and a couple of Vaxen.  The operating 
> system level code was robust; no crashes.
>
> We found some interesting bugs; I think my favorite was RSX wanting 
> all capital letters for months and Tops-20 sending capitalized months, 
> so file dates got messed up.  There were others, but they took some 
> effort to remediate; so, not so favorite...
>
> On 2/25/20 1:38 PM, Rob Jarratt wrote:
>
>     Mine has been built to be portable (it is in C). I have it running
>     on Windows, Debian and FreeBSD. I have not tried on VMS, but I did
>     do a partial port to VAXELN. I don’t foresee great difficulties
>     getting it to run on VMS. It looks like I may have done some
>     partial VMS work, but it has been a while since I looked. I could
>     look into reviving/continuing that if there is interest.
>
>     Mine won’t talk to TOPS10/20, yet. I started something on that too
>     if I remember correctly, I did build a DDCMP implementation which
>     I think partially worked. I think I foundered on just getting
>     DECnet running on TOPS20 on SIMH, I did some work with Mark
>     Pizzolato and Timothe Litt to build a KDP emulation, but I
>     struggled to keep up with them and gave up in the end. I am not
>     sure where that went.
>
>     I need a bit of time to work out what my status is I think as it
>     has been a while.
>
>     Regards
>
>     Rob
>
>     *From:*owner-hecnet at Update.UU.SE
>     <mailto:owner-hecnet at Update.UU.SE> <owner-hecnet at Update.UU.SE>
>     <mailto:owner-hecnet at Update.UU.SE> *On Behalf Of *Robert Armstrong
>     *Sent:* 25 February 2020 14:42
>     *To:* hecnet at Update.UU.SE <mailto:hecnet at Update.UU.SE>
>     *Subject:* [HECnet] Multinet alternatives ...
>
>        What are the alternatives for bridging DECnet systems across
>     the Internet that don’t require any special hardware and can be
>     self hosted on OpenVMS/VAX?
>
>       I’m aware of Paul’s Python router, but I didn’t think it ran on
>     a VAX.  Am I wrong about that?  Is there a Python for
>     OpenVMS/VAX?  I believe there’s a port for AXP but I didn’t know
>     about one for VAX.
>
>       And I saw Rob mentioned his user mode DECnet router.  Same
>     questions – what environment does that require?   And I saw
>     mention it being both Multinet and simh DDCMp compatible – will it
>     talk to a simh KS10/TOPS10/DMR ?
>
>     Bob
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.sonic.net/pipermail/hecnet-list/attachments/20200226/106e857d/attachment.html>