[Icecast-dev] patch for icecast-2.2.0 to add client maxtime (fwd)

[Geoff Shang]

Hi,

Sorry, I accidentally rejected this message instead of approving it.

Geoff.

---------- Forwarded message ----------
Date: Fri, 11 Mar 2005 17:33:06 +0100
From: Siegfried Wagner <esiggi at gmail.com>
Reply-To: siegfried at esiggi.net
To: icecast-dev at xiph.org
Subject: Re: [Icecast-dev] patch for icecast-2.2.0 to add client maxtime

Hi,

I did your recommended changes on my patch.
But with the "signed/unsigned" mismatch: That's also wrong with other
variables like source_timeout etc.

And I've got a fundamental question: Why are you kicking clients that
don't receive their data fast enough? Why not just throwing the data
away rather than terminating the whole connection!?
There would be missing parts or bad noises for the listener but at
least he could still be listening...

Another importing thing to notice is a problem with undetermined
crashes on FreeBSD with corrupted stacks so that there's no
possibility to do a debugging. The crash said segmentation fault or
bus error. It seems that it has something to do with the directory
servers because when they won't be used there were no crashes. I also
looked for a possible exploit but I couldn't see any strange
client-data when the crashes occured.

Siggi.


On Mon, 17 Jan 2005 15:06:38 -0800, Greg J. Ogonowski <greg at orban.com> wrote:
> Definitely agree this is a useful feature.
> -greg.
>
> At 14:50 2005-01-17, Michael Smith wrote:
>> On Mon, 17 Jan 2005 10:25:32 +0100, Siegfried Wagner
>> <siggi at radiofreaks.org> wrote:
>>> Hi,
>>>
>>> I wrote a little patch for the stable version (also works for the svn
>>> version) to add a new configuration parameter called
>> "client_maxtime".
>>> With this you can set a maximum connection time limit for a connected
>>> client so that you can disallow continuous listening.
>>> When the listening time exceeds the client connection will
>>> automatically be dropped.
>>> By default this feature is disabled (set to 0).
>>
>> This looks like a potentially useful feature. There are a number of
>> problems in your patch, though:
>>  - unsigned/signed confusion. client_maxtime should be unsigned
>> everywhere, not just in some places.
>>  - the check for whether to disconnect should probably be in
>> send_to_listener, rather than directly cluttering source_main.
>>  - The formatting is messed up, inconsistent with the style of the
>> rest of the code, and has several excessively long lines.
>>
>> If you could clean up those issues and send us another patch, then
>> (assuming others agree that this looks useful), we'd be happy to
>> include it.
>>
>> Mike
>> _______________________________________________
>> Icecast-dev mailing list
>> Icecast-dev at xiph.org
>> http://lists.xiph.org/mailman/listinfo/icecast-dev
>
> ________________________________________________________________________
> __
> Greg J. Ogonowski
> VP Product Development
> ORBAN / CRL, Inc.
> 1525 Alvarado St.
> San Leandro, CA  94577  USA
> TEL +1 510 351-3500
> FAX +1 510 351-0500
> greg at orban.com
> http://www.orban.com
>
>


-- 
| Dipl.-Inf. Siegfried Wagner
| LFE Medieninformatik (http://mimuc.de/)
| LMU München - Institut für Informatik
| Amalienstraße 17, 5.Stock, Zimmer 503, Tel: 089/2180-4655
-------------- next part --------------
A non-text attachment was scrubbed...
Name: icecast-2.2.0.diff
Type: application/octet-stream
Size: 6877 bytes
Desc: 
Url : http://lists.xiph.org/pipermail/icecast-dev/attachments/20050312/34e7de83/icecast-2.2.0.obj