Text 51, 151 rader
Skriven 2005-02-13 12:00:18 av Peter Knapper (3:772/1.10)
     Kommentar till text 50 av Mike Luther (1:117/3001.0)
Ärende: More on Sprint DSL?
===========================
Hi Mike,

Well I have read your message a few times and think I have some of it straight,
but I am still not quite sure...

 ML> What they THOUGHT they were going to do was put their 
 ML> new broadband ZyXel 324 router behind all this and 
 ML> isolate the OS/2 operation totally from everything else 
 ML> against the same ADSL line!  No I don't visualize any 
 ML> way to ever do that.  But maybe it could be done?  The 
 ML> ZyXel sure doesn't think so however I know to set that 
 ML> up.

No, you cannot connect 2 x ADSL devices to the same ADSL enabled telephone line
and expect both to work, ADSL is only a SINGLE device environment, it will work
with one device OR the other, but not both on the line together. Basically, you
cannot multi-drop ADSL, its point-to-point only.


 ML> Before I found this out, the new ZyXel 324 works as 
 ML> intended.  I can COMM port it, look at WAN and see 
 ML> traffic .. WAN lite winking and all that.  

What ADSL line did you connect it to? If the same line as the other ADSL device
then its likely its connection died...


 ML> But the OS/2 
 ML> box never got a connect to Sprint.  Of course not! So I 
 ML> know that at least the line given me through this multi-
 ML> story building goes somewhere and is live!

So I take that to mean the Zyxel WAN interface is configured correctly, but the
Ethernet side is not right somewhere.

 ML> So I tried to get a bare box single line DHCP connection with LAN0 set up 
 ML> for Netbios for OS/2 and TCP/IP protocols on it to 
 ML> simply connect to the provided cable that goes (?) to the closet.  No 
 ML> connect,  So my guess is that they are trying to stick 
 ML> this OS/2 box on the tail end of the Cisco switch and 
 ML> it isn't passing th OS/2 DHCP request properly to the 
 ML> Linksys in regard to some part of the protocol it needs. 

What model of Cisco is it? A Cisco Switch is a layer 2 device and unless
specifically configured not to, is totally DHCP transparent.

 ML> Even though as 
 ML> connected to the ZyXel router it gets a nice 
 ML> 192.168.1.33 private DHCP assignment just fine from it. 

Ok, so DHCP from the Zyxel works, but DHCP via the Cisco doesnt? That means the
DHCP server is not givng out Leases for some reason, or the PATH from the DHCP
Server to the OS/2 Client is being controlled in some way. 

To make sure my understanding is correct, your layout looks something like this
-

        ADSL Router with DHCP Server
         ||
        Cisco Switch
         ||    ||
         ||    ++==== DHCP to local machines that work fine.
         ||
         ++========== DHCP to your OS/2 machine that doesn't.

Then something is being done that is preventing the DCHP requests getting to
the ADSl Router or its replies from gettting back, OR you didn't actually get a
CONNECT to the Cisco Switch.


 ML>  Yes, I've read the Usegroup posting on having to 
 ML> enable additional options in the PROTOCOL.INI for the 
 ML> box in some cases for some devices.  But I'm not 
 ML> comfortable with doing that in the field yet from 
 ML> memory and wasn't prepared with even printed notes for 
 ML> this when I went to do a brand new single connect to a 
 ML> brand new Sprint ADSL modem!

If DHCP works for OS/2 from an ADSL Router directly connected, then this should
also work via another Etherswitch PROVIDED there is no specific configuration
in there blocking the DHCP traffic.

 ML>                                          
 ML> Since Peter here is Cisco oriented, maybe he can tell 
 ML> me what is missing from the OS/2 DHCP request protocol 
 ML> to get to the DNS service on the Linksys .. 

Nothing at all, DHCP should be transparent to a Cisco Switch.

HOWEVER you ned to be aware that -
  1. It is possible to configure a Cisco Switch Port to ignore and drop DHCP
traffic.
  2. If the Cisco Device is NOT just a simpoe Layer 2 Etherswitch but a layer 3
Router, then yes, things may be required to get DHCP traffic to pass.
If you could provide a diagram like above WITH Cisco Model numbers, and IP
Segment addressing info, then it may all become clear.


 ML> As I understand this, a proper way to isolate the entire OS/2 Peer 
 ML> operation for all the rest of the perfectly functional 
 ML> boxes in the OS/2 network is done through the use of 
 ML> two LAN NIC cards in the box.  

If you are only talking to OS/2 boxes on this ONE ethernet segment, then remove
NETBIOS over TCP/IP and you are secure for that segment and NETBIOS. If you
need to talk to Windows boxes on the same segment, then you will probably need
NETBIOS over TCP/IP and then you are vulnerable to all the Windows faults. 

Of course you could put a 2nd NIC in all the OS/2 boxes and create an isolated
LAN for them, but they only need NETBIOS and not NETBIS over TCP/IP on that
interface. Also ensure you do NOT Route beteen interfaces (IPGATE OFF). Of
course if the 2nd NIC only has NETBIOS and no TCP/IP at all, then routing is
non-existant anyway...


 ML> I can use MTPN setup and get the second NIC identified 
 ML> and assigned to protocols.  And .. as I understood 
 ML> this, only one of the NIC's may be set up with DHCP and 
 ML> not fixed addressing.  So my assumption is that 
 ML> whatever of the two NIC's I use for the DCHP connection 
 ML> to the whatever router Linksys or ZyXel, it is the 
 ML> other NIC which will have to connect back to the already present hub.

So you need to preserve the existing IP addressing for the OS2 boxes, but also
allow them to access the Router on a different Network address? Making a box
Dual homed (2 x NIC's), is one way, but I can see better ways. Without full
docs its hard to see what is being attempted and what is able to be changed.

 ML> When I go to log on to the LAN .. I get a message 
 ML> 'DOMAIN already in use' and as far as I can see .. that 
 ML> user is not even and has never logged in to anywhere.  
 ML> It, of course as PEER .. is IBMPEERS as far as I know.

This sounds like you have 2 x Netbios Domains defined and you are trying to
give them the same Name. Remove NETBIOS from one of the NIC's, unless you need
it on both sides, in which case you have an issue.

 ML> What am I doing wrong?   How do I enable this box so 
 ML> that the DCHP service can exist on the one NIC and 
 ML> IBMPEERS with the correct connections can be made to 
 ML> the local PEER LAN on the other NIC?

See above, if you dont need NETBIOS on a NIC, then remove it.

More info please.........;-)

Cheers...............pk.


--- Maximus/2 3.01
 * Origin: Another Good Point About OS/2 (3:772/1.10)