[pskmail] Re: jpskmail 0.3.8 beta test
- From: Rein Couperus <rein@xxxxxxxxxxxx>
- To: pskmail@xxxxxxxxxxxxx
- Date: Sun, 11 Oct 2009 16:31:42 +0200
John,
I will try the sever changes next week... I shot my test system at PI4TUE, and
it will not restart,
so I have to go there tomorrow...
Meanwhile I think I found the error in the client code, and repaired it. I have
tested it with some
quite big files, on a path with heavy multipath distortion, so it needed lots
of repeats.
The main problem was that the client generated block numbers outside the scope
of the
frame and put them into the $Missed variable. Since the change I have not seen
that happen again. The sliding window we use is 16 blocks, and the 'missed'
blocks
should be within that scope.
When generating the next frame, the server first sends the missed blocks, and
fills the
rest of the frame with new blocks, so we are not really missing anything,
unless the client generates
wrong numbers.
But I will try the server changes as soon as I have my test server back...
Meanwhile I have put a new server (0.9.20) with some important improvements
into the hermes.esrac.ele.tue.tue.nl/pskmail/alpha directory...
With that version the server will not get stuck in a deadlock when the client
does
not respond to connect acks...
Today we had some weird conditions on 30, lots of multipath giving the band a
'hollow' sound with
even some echo on it.... PSK250 was really struggling, so ideal for testing :)
THOR22 always worked...
73,
Rein PA0R
> -----Ursprüngliche Nachricht-----
> Von: "John Douyere" <vk2eta@xxxxxxxxx>
> Gesendet: 11.10.09 11:59:29
> An: pskmail@xxxxxxxxxxxxx
> Betreff: [pskmail] Re: jpskmail 0.3.8 beta test
Rein,
>
> On the re-send of missing blocks issue, without testing, just by
> looking at the server's code, may I suggest the following changes in "
> sub handle_rxqueue":
>
> 1. move the "$cntruns++;" statement in the else part of the test for
> good frames. The logic being is that we would stop after having found
> 8 missing blocks, not just after checking 8 positions for missing
> blocks. The current code is consistent with the behaviour I found:
> only the first few missing blocks are requested, not all (or at least
> up to 8) of the missing blocks.
>
> 2. I think we should remove the section below, since it is possible
> to have missed 8 blocks, but received the status block OK, or have
> accumulated 8 missing blocks. The test in 1 above would in any case
> limit the number of missing blocks (there is also a missing string
> truncation for that purpose):
>
> if (length ($MissString) == 8 && $Iamserver) {
> $MissString = "";
> $HisLastblock = $Goodblock;
> }
>
> Sorry but I don't have time to test that code.
>
> Regards,
>
> john
>
> On Sun, Oct 11, 2009 at 8:38 PM, Rein Couperus <rein@xxxxxxxxxxxx>
> wrote:
> I noticed myself yesterday when watching pa0son test with pa0son-4,
> I am on it now...
>
> Rein PA0R
>
>
>
--
http://pa0r.blogspirit.com
Other related posts:
