OS/2 Wireless Users Mailing List <os2-wireless_users@2rosenthals.com>
Christian Langanke wrote:
[...]
> I am sorry, but having two interfaces configured to the same subnet is
> wrong. If you mean that you disable one of either interface, this would
> be ok. But it is definitely not sufficient to just "not use" one of
> either interface. The IP stack will not care for if you don't use one of
> two interfaces - exactly one of two will just always win and the other
> will never be used.
It's absolutely no problem to have two interfaces in the same subnet as
long as:
+) There is only one route to the subnet
+) The hostid is set correctly
>> +) Even if it is necessary to delete one of the interfaces it makes the
>> TCP/IP-stack unusable because if it turns off the previously configured
>> interface it does not update the routes - so no interface does work:
>>
> That is correct and I hope I will be able to fix this route issue.
Good!
[...]
> Hmmm. I did not know that this has to be explicitely set. How would I do
> that ?
'@ifconfig' interface '| find "inet" | rxqueue'
if queued() > 0 then parse pull . 'inet' ip .
else ip = 'ERROR'
return ip
>> 9) Run rfcaddr for TCPBEUI
>>
> Note that NetBEUI is currently ot handled by XWLAN, so you would have to
> do that with the script feature.
I know, but the problem is that the script is run before the WLAN
interface is configured (which I think is OK). And rfcaddr.exe should be
run afterwards.
> I would be interested though to have an exeperienced user telling me
> what else could be required/make sense for better NetBEUI support.
During my years with TCPBEUI I never could find out when it is really
necessary to run rfcaddr.exe and when not. Or: when a multi interface
configuration does work properly in all cicumstances. Either it is a big
miracle, OS/2's implementation of TCPBEUI is lousy or (most likely) I am
just to dumb.
[...]
> For testing I implemented something different, at least for the default
> route. If the WLAN interfaces is to be kept, the default route is saved,
> deleted, then LAN is deleted, then default route is reset - et voila:
> the defalt route is attached to the WLAN interface. I intend to cache
> _all_ routes for the cabled interface and restore them all on the
> correct interface, so network and hostroutes would also be handled
> properly. The only drawback is that I cannot do that for the way back
> (wireless to cabled). Here the setup script is to setup all routes
> properly again.
That sounds OK to me. But it should be OK just to make sure that the
route to the subnet uses the wLAN interface and the default route is
restored as you mentioned above. In most environments all other routes
are just left overs from previous connections which transparently get
"rebuilt".
[...]
> I am sorry to hear that. Did you post a request to netlabs.org on this ?
No, it's the IBM driver - don't know if one at netlabs could help me out
here. Just for information: Some time ago I checked everything: And I
mean everything: Modifying each and every value which can be configured
in PROTOCOL.INI and in various access points.
But even in the Wind***'s world it's not much better: If you use a
driver that's to new then after installing SP2 IBM's Access Connection
does not work with the card anymore.
Tanks anyway (for listening) ;-)
Michael
--
Michael Warmuth, Manager IT, michael@warmuth.at
Michael Warmuth - EDV-Dienstleistungen, www.warmuth.at
Waldegghofgasse 27, A-1170 Wien, Austria - EUROPE
Tel: +43-676-3863440, Fax: +43-1-4892728-99
Austria -- The place in the heart of Europe
where no kangaroos are hopping around.