Balazs, While I kinda like nak's (Expert User's) idea, it may be easier to do this:
I would suggest that you run a basic IP routing daemon. Remove the default routes from the interfaces, and enter two static, default routes in the routing daemon's config. With most modern implementations of IP routing, you can assign a "cost" to a route. Set the "cost" of the default route for the wireless interface to be higher than the cost of the 100bT interface, and your machine will use the 100bT when it's up, and the wireless link when it's not. This does depend on the interface between the cards and the network drivers taking the interface up and down when link-state changes, so the routing daemon can invalidate the links, or to your local router being willing to exchange routing information with you. It's been a while since I've done this on a unix system. It was routed (as in route-d, or routing daemon) in my day, for RIP1 routing. I don't know what the daemons are for RIP2 and OSPF routing . . . . a quick search on the debian packages in stable show bird and zebra as routing daemons you could use. Multihoming systems on the same IP subnet has been a headache for years, and usually requires extra software to work properly. As an example, here's how the Windoze server world handles it out of the box (NT 4 and earliers were a total disaster - the Win2k Server scenario presented here is a VAST improvement, and it still sucks, and it only deals with services OUTBOUND from the server, not inbound services). OK - you have a win2k box with 2 ethernet cards, one 100mb and the other 10mb, plugged into different ports on a switch network. They each have a unique IP. Win2K doesn't associate a default gateway with an interface, but since the default gateway is on the same network as both cards, two default routes are created in the routing table, one for each interface. OK - so it boots, and happens to pick the 100mb interface first, and uses the 100mb interface for the default route. The 2nd interface is up, and connected, but it won't be used. OK - now the 100mb interface loses it's connection, win2k disconnects it, the default route expires, and the 2nd default route is now the only default route and everything starts heading out that interface. As long as the server your client application was talking to doesn't mind the change in source IP, you're fine. For examply, you were telnetted into a host - after the 100mb interface goes down, the telnet packets will be sent with the 2nd interface's IP as the source address. Your host has no clue that this is the same machine, and ignores them . . .. OK - assuming you can live with that, what happens when the 100mb interface comes back up? Well, in the Win2K world, you will continue to use the 2nd, 10mb interface until it stops working, and only then will you switch back. Good luck, and let me know how this works out for you . . . madmac On 30 Oct 2002, 22:05:42, Balazs Javor wrote: > > > Hi, > > Thank you all for your suggestions. > They basically confirm my fears that it isn't going to be > very simple. > I must admit I'm not a networking expert, allthough I thought, > that I have a good understanding of the basics. > Please let me give you some more details on my (probably typical) > setup and some of my goals/assumptions. Maybe you could then > suggest me the best way to proceed: > > I have an ADSL router connected to a network switch along with > 3 PCs connected to the same switch. They all are on the same subnet > and all have the router as the default gateway. So long so fine. > They can all talk to each other, share files and access the Internet. > > Now comes the notebook. I have configured (for now) the built-in NIC > the same way as with the other machines, so it works as well. > Now, I've ordered the Orinoco Gold PCMCIA card along with the RG-1100 > wireless access point. It is my understanding that you can set up the > access point to act purely as a bridge. Where my assumption was that > it will basically act as a "wire to radio signal converter" and if it > were not for the built-in NIC, I could just plug the access point into > the switch, the PCMCIA card into the notebook and configure the card to > be on the same subnet as the other machines, and everything would > work as before. > > And if this is the only easy way, then I would simply not use the > built-in NIC. However, since the built-in NIC is about 10 times > faster, I was kinda hoping that I could somehow make it possible, > that whenever I am at my desk or whenever I need to transfer large > amount of data to the notebook I could just simply "switch" to the > built-in NIC. > Now if I need to configure either of the cards to be on a separate > subnet I would need to route that subnet to the original network > somehow as both the fileserver and the Internet gateway are on that one. > So I would need to set up one of the other machines with another NIC > and set it up to be the router between the two networks. And this I > think is a bit too complicated for what I wanted to use it for... > > However maybe it is possible to have both NICs be configured for the > same IP address and manually switch between them, so that only > one of them is up at a given moment? > > So what do you think is the best approach? > Many thanks again for your help! > regards, > Balazs > > -- Doug MacFarlane [EMAIL PROTECTED] -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
