[mosh-devel] Lightning Talk in German about Mosh and AutoSSH

[Keith Winstein]

Axel:

Awesome, thanks for your interest and your kind words! Would certainly
be interested in the English slides.

You may have seen Mosh hit Hacker News today
(http://news.ycombinator.com/item?id=3819382) -- I wonder if somebody
from your talk tipped off the submitter?

My easy technique these days for getting a large network delay is just
to run something like:

$ cat /dev/zero | ssh remotehost 'cat > /dev/null'

That seems to do a good job of filling up the network buffers and
induces a >1 second delay sometimes.

There are more sophisticated approaches if that doesn't work. :-)

Best regards,
Keith

2012/4/9 Axel Beckert <abe at deuxchevaux.org>:
> Hi,
>
> JFYI: I gave a lightning talk (10min + questions) in German about SSH
> and unreliable network connections yesterday at Chaos Computer Club's
> annual Easterhegg event:
>
> https://easterhegg.ch/lecture-ssh-ber-unzuverlssige-leitungen-lag-roaming-gsm-wackelige-wlans-und-hibernation
>
> I presented AutoSSH and Mosh as possible tools to workaround those
> network issues which are typically annoying to SSH users. There were
> quite a lot of questions about mosh and I redirected those questions I
> couldn't answer by mind to the new mosh brand-new website.
>
> Unfortunately I hadn't yet read the SSP section of mosh's website when
> preparing the talk, but I read it now and will mention it in future
> talks.
>
> The slides (S5/HTML) of the talk are available at
> http://noone.org/talks/ssh-tricks/ssh-tricks-eh12.html
>
> I likely give that talk again in two weeks at the local LUG and
> possibly also at other events this year.
>
> If wanted, I can post a short notice when I have slides in English
> available, too.
>
> Oh, and some anecdote from the talk: Of couse I tried to show the
> output prediction, but the network was either too bad (WLAN, didn't
> get an IP at all during preparing for the talk, worked fine during the
> talk) or too good (LAN, everything worked fine). So during the
> preparation I tried to artificially lower the bandwidth with trickle,
> but even with values around 1 byte per second, typing was still
> fluently so I suspect trickle's library preloading didn't really catch
> mosh's network system calls. The only way I got some artificial lag
> was using the iodine IP over DNS tunnel, but I didn't do that during
> the talk. So I just presented how Mosh way more quickly reestablished
> the connection after an IP change compared to AutoSSH.
>
> So any hint on how to get some artificial bandwith or latency
> limitation for mosh to show off the output predicition would be nice.
> :-)
>
>                Kind regards, Axel
> --
> /~\  Plain Text Ribbon Campaign                   | Axel Beckert
> \ /  Say No to HTML in E-Mail and News            | abe at deuxchevaux.org  (Mail)
>  X   See http://www.asciiribbon.org/              | abe at noone.org (Mail+Jabber)
> / \  I love long mails: http://email.is-not-s.ms/ | http://noone.org/abe/ (Web)
> _______________________________________________
> mosh-devel mailing list
> mosh-devel at mit.edu
> http://mailman.mit.edu/mailman/listinfo/mosh-devel