[mosh-devel] classification

[Keith Winstein]

Hi Dave,

I took your commit and put it on my master branch (now pushed), so
people should feel welcome to try that out. It will also be in the
mosh-dev daily Ubuntu PPA if anybody is brave enough to run that
(ppa:keithw/mosh-dev).

I have also deployed this commit on the server many people use at MIT,
so hopefully we will get some good testing over a variety of Internet
paths. I am a little concerned about whether ECT will pass
successfully over wide-area paths (given the FUD about ECN in
general), but I guess we just have to find out.

Do you think we are being bad citizens by advertising ECT but not
actually doing anything in response to the marking? Right now, the
only way to get Mosh to speed up or slow down the frame rate is to
change the SRTT. We do not respond to losses at all. We just aim to
send no more than one Instruction per SRTT/2 (i.e., have at most one
Instruction in flight at a time on average).

Best regards,
Keith

On Fri, May 25, 2012 at 8:20 PM, Dave Taht <dave.taht at gmail.com> wrote:
> could I encourage folks to try diffserv marking and maybe ECN?
>
> In my branch of the  repo, I'm using AF42, which has the proper bits
> to end up in the VI queue on linux wireless, and is mildly prioritized
> for other stuff.
>
> Setting ECN capability on ensures that new AQMs such as fq_codel
> will mark, rather than drop packets. I do note it would be
> good to evolve an ECN (re) negotiation protocol for when users
> move from network to network.
>
> On Fri, May 25, 2012 at 11:33 PM, Keith Winstein <keithw at mit.edu> wrote:
>> Hello Mosh users and developers,
>>
>> mosh 1.2.1 has been released.
>>
>> The source code is at:
>> https://github.com/downloads/keithw/mosh/mosh-1.2.1.tar.gz
>>
>> This release fixes a number of issues with mosh 1.2:
>>
>> * Ignore nonsensical escape sequences. Timo Juhani Lindfors found that
>> mosh faithfully interprets even very large "repeat" counts in ANSI
>> escape sequences, allowing a malicious application to cause the
>> mosh-server to use a lot of CPU time trying to execute a short ANSI
>> escape sequence. The same sequences can allow a malicious mosh-server
>> to cause the mosh-client to use a lot of CPU time. Mosh 1.2.1 ignores
>> these large repeat counts. This issue has been tagged as
>> CVE-2012-2385.
>>
>> * Improve performance on lossy links
>>
>> * Give the user a helpful diagnostic when the link is dead in only one
>>  direction
>>
>> * Use less CPU when link is down (Keegan McAllister)
>>
>> * Use less memory when mosh-server is malicious.
>>
>> * Fix a vttest regression re: wrapping and tabs.
>>
>> * Enable a roundtrip verifier of terminal emulator correctness when
>>  the server is verbose.
>>
>> * Remove skalibs as a dependency (Keegan McAllister)
>>
>> * Remove use of poll() and the OS X poll workaround in favor of
>>  pselect(), which we think works everywhere (Keegan McAllister)
>>
>> * Include a bash_completion file (ejeffrey)
>>
>> * Include a firewall profile for UFW (Fumihito YOSHIDA)
>>
>> * Warning on out-of-order/duplicated datagrams (or failed nonce increment).
>>
>> * Clearer error message on invalid port number.
>>
>> * Cleanups to quit scenario when firewalled.
>>
>> ===
>>
>> mosh 1.2.1 is backwards-compatible with mosh clients back to 0.96 and
>> mosh servers back to 1.0.9. Please let us know of any problems
>> (https://github.com/keithw/mosh/issues).
>>
>> Best regards from the Mosh team,
>> Keith
>>
>> On Thu, Apr 26, 2012 at 3:06 AM, Keith Winstein <keithw at mit.edu> wrote:
>>> Hello Mosh users and developers,
>>>
>>> mosh 1.2 has been released.
>>>
>>> The source code is at: https://github.com/downloads/keithw/mosh/mosh-1.2.tar.gz
>>>
>>> This version polishes a number of rough edges, and especially:
>>>
>>> * Boost is no longer required to compile Mosh
>>>
>>> * Mosh passes the locale over the connection for users who don't have
>>> SSH configured to do so
>>>
>>> * Mosh provides more helpful diagnostics
>>>
>>> * The client has a --ssh= option to give options to the startup SSH,
>>> including setting a port number or key file
>>>
>>> * Mosh uses compiler hardening flags when available
>>>
>>> * Added support for FreeBSD, Cygwin, RHEL/CentOS 5, OS X 10.5 on PPC
>>>
>>> * Improved support for OS X 10.6 and ARM platforms.
>>>
>>> * Many bug fixes and some performance improvements
>>>
>>> ===
>>>
>>> The new version is backwards-compatible with mosh clients back to 0.96
>>> and mosh servers back to 1.0.9, but to realize some of these benefits
>>> you will need version 1.2 installed on both client and server. Please
>>> let us know of any problems (https://github.com/keithw/mosh/issues).
>>>
>>> Best regards from the Mosh team,
>>> Keith
>> _______________________________________________
>> mosh-devel mailing list
>> mosh-devel at mit.edu
>> http://mailman.mit.edu/mailman/listinfo/mosh-devel
>
>
>
> --
> Dave Täht
> SKYPE: davetaht
> US Tel: 1-239-829-5608
> http://www.bufferbloat.net