[JDEV] jabberd patch
Richard Dobson
richard at dobson-i.net
Tue Feb 25 18:15:09 CST 2003
Ah well there you go then, I was just getting at the fact that there is
a way for clients to tell and that they must be fixed to only
auto-reconnect on network problems.
On Wednesday, February 26, 2003, at 12:02 am, David Waite wrote:
> Both the open-source jabber.org and the commercial jabber.com
> implementations send a stream:error before replacing the current
> session with a new one. Exodus is an example of a client which does
> not reconnect in this case.
>
> -David Waite
>
> Richard Dobson wrote:
>
>> Exactly but the thing is that they dont get a stream error, at the
>> moment the server ends the session with a </stream:stream> and drops
>> the socket, and because that sequence of events means the server is
>> terminating the connection, which i think will only happen at the
>> moment because of another session connecting, that can be used for
>> clients to determine wether they should try auto-reconnecting,
>> because IMO clients should only be auto-reconnecting in the case of a
>> network problem terminating their connection and in that case they
>> would not have received the </stream:stream> from the server so that
>> is why you can use that to determine wether to auto-reconnect.
>>
>> On Tuesday, February 25, 2003, at 09:12 pm, David Waite wrote:
>>
>>> IMHO, this is a bug in those clients. If the server says it is
>>> disconnecting you (via a stream:error), you should not immediately
>>> reconnect.
>>>
>>> -David Waite
>>>
>>> Tijl Houtbeckers wrote:
>>>
>>>> "Richard Dobson" <richard at dobson-i.net> wrote on 25-2-2003 13:31:12:
>>>>
>>>>> Yea sure I wasnt arguing against the "option", but it shouldnt be
>>>>> standard behaviour and must be documented that it can cause
>>>>> issues/problems, so the admin can make an informed choice of which
>>>>> behaviour they want.
>>>>>
>>>>
>>>> Then again, not using this patch has it's own problems when two
>>>> clients with auto-reconnect start to "fight" over the same
>>>> connection. So this patch is definatly usefull for those who want
>>>> to avoid this type of behaviour.
>>>>
>>>
>>> _______________________________________________
>>> jdev mailing list
>>> jdev at jabber.org
>>> http://mailman.jabber.org/listinfo/jdev
>>
>>
>>
>> _______________________________________________
>> jdev mailing list
>> jdev at jabber.org
>> http://mailman.jabber.org/listinfo/jdev
>
>
> _______________________________________________
> jdev mailing list
> jdev at jabber.org
> http://mailman.jabber.org/listinfo/jdev
More information about the JDev
mailing list