ok, but what about adding a similar implementation as mpm-itk uses:
simulate timeout and close connection so browser would reconnect. It is 
said to work well in practice and to be compliant with RFC2616


Leen Besselink wrote:
> Janno Sannik zei:
>   
>> We would like to use Keepalive with peruser and possibly fix it if
>> needed, but we can't seem to find a problem (by just roughly testing it)
>> why it shouldn't work.
>>
>> Could someone explain us a little how to reproduce possible security
>> breach or some other malfunction then using keepalive. We would gladly
>> try to produce a workaround.
>> With simple test using Keepalive on and using 5 domains no problems seem
>> to occur. All requests get handled correctly.  Also apache server-status
>> shows 5 keepalive connections. One for each vhost.
>>
>>     
>
> Hi Janno,
>
> The problem is kinda by design.
>
> It can't really handle this situation:
>
> If a HTTP-client first sends a host-header for one vhost after that
> sending an other request over the same connection for an other vhost.
>
> Because the socket-descriptor is passed to the vhost-child-process,
> but it doesn't get passed back to the parent process.
>
> Obviously that would also be the solution, but the question is, would
> this then still be secure.
>
> You can probably work around the problem with having each vhost on a
> seperate IP-address, the browser would know it's talking to the same
> server.
>
> PS If I remember correctly, because this is what it is for the metux-mpm.
>
>   
>> Janno Sannik
>>
>> _______________________________________________
>> Peruser mailing list
>> [email protected]
>> http://www.telana.com/mailman/listinfo/peruser
>>
>>     
>
>
> _____________________________________
> New things are always on the horizon.
>
>
> _______________________________________________
> Peruser mailing list
> [email protected]
> http://www.telana.com/mailman/listinfo/peruser
>   

_______________________________________________
Peruser mailing list
[email protected]
http://www.telana.com/mailman/listinfo/peruser

Reply via email to