Anthony DeRobertis schreef:
On Fri, Sep 21, 2007 at 06:16:13PM +0200, Rob Janssen wrote:
This is of course a bug in Windows XP. Would you think that someone at Microsoft actually notices this and will do something about it? Apparently not.

This isn't a bug; it's the intended behavior. See
<http://www.microsoft.com/technet/technetmag/issues/2007/09/CableGuy/>

The NTP client assumably picks an address when it starts and sticks with
it as long as possible; this makes sense --- jumping between networks
would cause jitter, why would it want to do that? Restarting the NTP
service might help.
It could be intentional, but still it isn't going to work. So I insist on calling it a bug, while others might call it a feature.

The article explains a couple of generic concepts which are sound, also when related to the current discussion and your remarks above, but in the special case of a tunnel connecting private-address networks over the public internet, sending packets with a public source address to a private tunnel endpoint just doesn't make sense. Either the packet or its reply cannot be delivered, so there will be no useful communication.

It is arbitrary if the NTP service should change the packet source address or the IP layer should simply drop the packets, but as it is now it is never going to work.

But my main point was to show that tracing a network and revealing packets to or from a private address space is not that unusual, can be the result of a well-meant setup that fails because of some bug or feature, and that this condition does not mean that a problem must exist for the end-user that he/she is surely going to troubleshoot and solve. The user is happily using the laptop, unaware of the funny NTP packets the system is emitting. (once every 16 seconds... Microsofts implementation of NTP has the usual mistake when handling nonresponding NTP servers: it keeps polling at the highest rate)

Rob
_______________________________________________
timekeepers mailing list
[email protected]
https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers

Reply via email to