[nos-bbs] A patch to JNOS2.0f is released

[Jay Nugent]

Greetings Skip,
   I was expecting all you were changing was the setting of IRTT and
wondered why recompile the NOS kernel when all that was needed was to
change the 'tcp irtt <nnnn>' value in autoexec.nos???  Then I found your
note (not at all mentioned in your email) saying that you *changed* the
FORMULA for calculating MDEV:

   "The second patch changes the time-to-wait formula as and optional 
    feature. The normal formula uses MDEV variable as a "smoothed" 
    deviation, the modified formula uses MDEV as a "high-water-mark" of 
    experience. The new formula reacts faster to slow connections. It 
    performs exceptionally at the authors site, and is offered to the 
    community for evaluation."


   I wonder if anyone has yet tried out this patch, and if so, does it 
improve or worsen the conditions on a BUSY link???  It may very well fix 
problems on slow links with two or three stations, but it needs to work in 
both worlds to be usefull.  Has anyone thried this yet???

   ...more comments embedded in-line...


On Fri, 27 Jun 2008, (Skip) K8RRA wrote:

> Greetings.
> 
> The Beta testing has been successful, so it is time to make the "bb"
> patch available for JNOS.  It addresses the TCP protocol use of historic
> measurements of round-trip-time [RTT] for packet acknowledgment and
> improves the performance of re-transmittal protocol.
> 
> The short background is:
> I found my station buried beneath huge quantities of unneeded repeat
> packets -- TCP was badly deciding that packets were lost.  The work
> following discovery led to two changes that have improved life here:

   If you run the LINK layer in connected mode it will perform forward
error correction (FEC) and provide RETRIES.  And the PACKET layer (IP)
*also* providing FEC, so it will *also* sends retries for the very same
failure!  Shouldn't run BOTH -- especially on a marginal RF path.  You may 
not ever notice it on a good RF path, however.

   So my first question -- are you running IP over Connected Mode??

   Secondly -- let us see your Global AX.25 and TCP settings.  Then we can
determine if you may have missed a setting that would have eliminated the
abundance of retries you were seeing *without* having to modify the source
code and recompile.  The need to recompile always scares me...
 

 
> First, former session measurements of RTT were being lost and thus IRTT
> was being applied too often resulting in the need to re-learn the
> characteristics of the connection.
> 
> Second, the time-to-wait formula was changed and IRTT was used to also
> set MDEV factor which improved accuracy and shortened learning time.
> This change may be switched on & off.

   Oh, I guess you DID mention the formula change.  I just didn't catch it 
when I read your message the first time :(

 
> The details in this work are found in the "readme-bb" file that is part
> of the released material.  My code base is Maiko's jnos2.0f, so direct
> module substitution works.  There is no conflict with the recent 2.0f2
> patch, I am running both.  This patch does contain a modified config.h
> so some editing on that file for local conditions is a must.
> 
> If your code base is an earlier jnos, I expect direct substitution is
> not applicable.  I did note that the area of my changes seem to go way
> back in history, so it may be straight forward to apply the change to
> earlier jnos revisions thru editing rather than substitution.
> 
> Access to this patch is thru the wiki at http://jnoswiki.no-ip.org/ then
> follow the path: SoftwareInstall -- PATCHES TO SOFTWARE -- SoftwarePatch
> -- Attachments.  Some of the surrounding text may also be useful.
> 
> As with all GPL software, expect the unexpected.  Let me [us] know if
> this causes grief.
> 
> 73
> de [George (Skip) VerDuin] K8RRA k


   Thanks!
      --- Jay Nugent  WB8TKL

In response to the issue about dumbing-down Amateur Radio, a fellow Ham 
said:    "Would you like fries with that emergency communication?"
+------------------------------------------------------------------------+
| Jay Nugent   jjn at nuge.com    (734)484-5105    (734)544-4326/Fax        |
| Nugent Telecommunications  [www.nuge.com]     (734)649-0850/Cell       |
|   Internet Consulting/Linux SysAdmin/Engineering & Design/ISP Reseller |
| ISP Monitoring [www.ispmonitor.net] ISP & Modem Performance Monitoring |
| Web-Pegasus    [www.webpegasus.com] Web Hosting/DNS Hosting/Shell Accts|
| LinuxNIC, Inc. [www.linuxnic.net]   Registrar of the .linux TLD        |
+------------------------------------------------------------------------+
  1:01am  up 251 days,  1:52,  5 users,  load average: 0.10, 0.17, 0.17