On 2014-01-16 09:18, Bob Armstrong wrote:
Multinet is horribly unreliable because it is based
on a fundamentally defective protocol design.
Really ? I've been using it for years and haven't noticed this.
Multinet tunnels work just fine most of the time, but Paul is right in that they do abuse things.
The HECnet bridge, OTOH, crashes regularly. It doesn't seem to recover
gracefully from network hiccups.
Huh? Since it's totally stateless that is weird. Let me know if you actually want me to look at it. Also, I've never had it crash, but the code is far from pretty, and depending on compiler and environment, I can believe that there are subtle bugs to be found.
Don't have a Cisco, so I can't comment on that.
:-)
Johnny
Hello!
Before we completely beat Multinet down with cudgels and assorted
sharp things, I have a question or two:
For one thing, Eisner is who? (What specialty does this machine do?)
And for another how did the machine in question get named?
I can cite complex examples behind why my machines get their names,
its almost a definite cause to change the subject.
Once those are answered we can go back to beating that protocol model
with those things.
-----
Gregg C Levine gregg.drwho8 at gmail.com
"This signature fought the Time Wars, time and again."
On Thu, Jan 16, 2014 at 12:36 PM, Bob Armstrong <bob at jfcl.com> wrote:
Then you're very lucky. I had nothing but trouble.
There are quite a few people on HECnet using Multinet - anybody else
besides Paul have problems?
Bob
Then you're very lucky. I had nothing but trouble.
There are quite a few people on HECnet using Multinet - anybody else
besides Paul have problems?
Bob
On Jan 16, 2014, at 12:18 PM, Bob Armstrong <bob at jfcl.com> wrote:
Multinet is horribly unreliable because it is based
on a fundamentally defective protocol design.
Really ? I've been using it for years and haven't noticed this.
Then you re very lucky. I had nothing but trouble.
The reason is that Multinet uses the point to point mode of DECnet but over UDP. That clearly and explicitly violates several of the requirements of the DECnet routing architecture specification.
The two main trouble areas are (a) packet loss UDP has it, point to point DNA datalinks do not; (b) lack of restart notification so if one side times out or otherwise restarts for some reason, there is a lot of back and forth before things finally sort out.
GRE has none of this, because it uses the Ethernet datalink mode, which is the correct one to use for a datagram connection.
In other words, the people who designed GRE tunneling understood DECnet, while the ones who created Multinet tunneling were either clueless or just didn t care that they were doing it wrong.
paul
Multinet is horribly unreliable because it is based
on a fundamentally defective protocol design.
Really ? I've been using it for years and haven't noticed this.
The HECnet bridge, OTOH, crashes regularly. It doesn't seem to recover
gracefully from network hiccups.
Don't have a Cisco, so I can't comment on that.
Bob
Sorry if it wasn't clear... Two different threads had a little
cross-posting I think. Here I'm just trying to establish a solid
HECnet peer. Johnny-bridge would have been preferable since I'm
running this on my NAS (and I know 'bridge' works), but if a GRE peer
is easier to find, by all means. While my QNAP NAS does have several
tunneling options in its stock config (it's linux arm), it appears
they did not include the ip_gre module, so I'll have to see about
hacking that in. Assuming that goes well, who's solidly offering the
GRE tunnel? We can exchange IP's and configs off-list.
Thanks in advance,
Mark
On Wed, Jan 15, 2014 at 11:25 PM, Dave McGuire <mcguire at neurotica.com> wrote:
I thought you and I were only discussing some X.25 experimentation.
If you want a full-on HECnet connection via GRE, Brian Hechinger can
stick your info in his database, and tunnel configurations will be
automatically generated for all of us on his next run.
-Dave
On 01/16/2014 02:12 AM, Mark Abene wrote:
Seems rather counter to the idea of HECnet if it's this difficult to
join it.
On Jan 14, 2014 10:13 PM, "Johnny Billquist" <bqt at softjar.se
<mailto:bqt at softjar.se>> wrote:
There are definitely people running the bridge in the US, but if
they want to peer is another story.
Johnny
On 2014-01-14 21:31, Mark Abene wrote:
I'm going to try getting a GRE tunnel going with Dave McGuire
soonish... In the meantime, if someone has a speedy connection
and is
already running 'bridge' (which would be simplest) let me know.
Thanks,
Mark
On Tue, Jan 14, 2014 at 5:25 PM, <Paul_Koning at dell.com
<mailto:Paul_Koning at dell.com>> wrote:
On Jan 14, 2014, at 8:17 PM, Mark Abene <phiber at phiber.com
<mailto:phiber at phiber.com>> wrote:
I'm not using the virtual cisco for HECnet at all,
that's only for the
X.25 project (currently at sampsa.com
<http://sampsa.com>, though I may run a local node
additionally).
For HECnet I'm just running a plain old fashioned
Johnny-bridge. :)
I take it HECnet peers are lacking in the U.S.?
I guess I should put my Python router up permanently...
paul
--
Dave McGuire, AK4HZ
New Kensington, PA
<Paul_Koning at Dell.com> writes:
On Jan 16, 2014, at 8:22 AM, Brian Schenkenberger, VAXman- <system at TMESIS.C=
OM> wrote:
Johnny Billquist <bqt at softjar.se> writes:
=20
...
If they are running Multinet, you could also setup a link using that.
=20
Oh yes! Eisner is or was running Multinet.=20
If you have a choice between Multinet and GRE, use GRE. Multinet is horrib=
ly unreliable because it is based on a fundamentally defective protocol des=
ign.
Absolutely.
I'll know more once the eagle has landed. ;)
--
VAXman- A Bored Certified VMS Kernel Mode Hacker VAXman(at)TMESIS(dot)ORG
Well I speak to machines with the voice of humanity.
On Jan 16, 2014, at 8:22 AM, Brian Schenkenberger, VAXman- <system at TMESIS.COM> wrote:
Johnny Billquist <bqt at softjar.se> writes:
...
If they are running Multinet, you could also setup a link using that.
Oh yes! Eisner is or was running Multinet.
If you have a choice between Multinet and GRE, use GRE. Multinet is horribly unreliable because it is based on a fundamentally defective protocol design.
paul
Mark Wickens <mark at wickensonline.co.uk> writes:
It *almost* makes me want to get the missus to drape herself over an=20
Alphaserver, or such like... ;)
Maybe all this talk of line printers should spur someone to make the DEC=20
equivalent of Benny Benassi's 'Satisfaction' video:=20
https://www.youtube.com/watch?v=3DV5bYDhZBFLA
Mark
Well, I already have a soundtrack. It just needs the hot female visuals:
http://tmesis.com/soundz/IWantYourVAX.wavhttp://tmesis.com/soundz/IWantYourVAX-Lyrics.txt
--
VAXman- A Bored Certified VMS Kernel Mode Hacker VAXman(at)TMESIS(dot)ORG
Well I speak to machines with the voice of humanity.
