[hipl-users] Re: HIP timers?
- From: Simon Schuetz <simon.schuetz@xxxxxxxxxxxxx>
- To: hipl-users@xxxxxxxxxxxxx
- Date: Mon, 05 Apr 2004 11:28:11 +0200
Readdressing definitely fails, the stationary node does not update the
HIP-IP mapping in
those cases were the test fails.
My analysis of TCP timeout algorithms says that the retransmission timers
are bound to
disconnection time, but should not be affected by the time the connection
was up *before*
disconnection.
What do you mean by "The first packets are not cached either" (which
packets, HIP or data ...)?
Simon
On Sun, 4 Apr 2004 15:50:30 +0300 (EEST), Miika Komu <miika@xxxxxx> wrote:
> On Fri, 2 Apr 2004, Simon Schuetz wrote:
>
>> When t2 (disconnection time) is about 100 seconds, then the tests mostly
>> fail if t1 (first connection time) is lower than 14 seconds. No data is
>> sent after the reconnect. Only some HIP REAs are sent, no actual data
>> packets.
>> If however t2=100 and t1>16 seconds, the tests mainly succeed!!!
>>
>> Are there any timers in the HIP code that could produce these results?
>> I guess the readressing fails for some reason. It shouldn't be a matter
>> of TCP, since the time of disconnection doesn't change. I'll do some
>> testings with UDP streams, just to be sure, soon.
>
> There are no timers implemented yet in the code (even though there should
> be). The first packets are not cached either. It is up to the TCP to
> retransmit packets.
>
> Hmm, this is wierd... or perhaps not. An analysis of TCP timeout
> algorithm
> and window sizes could help in understanding this behaviour.
>
> You can check if the readdressing fails from
> "/proc/net/hip/sdb_peer_address". The last address in the list is the
> default peer address.
>
Other related posts:
