On Mon, 11 Feb 2008 11:28:16 -0600 [EMAIL PROTECTED] wrote: > I happen to be in a situation where I have both a DSL and CABLE > connection to internet up for the time being... (Until the DSL > contract month runs out). > > It affords a nifty opportunity to do some experiments. Of course I > tested the speeds of both and it varies between 200 and 500 % faster > on the Cable connection. (Nice).
Neat. You should set up advanced routing so you can use both at once. > At first I used single machines connected independently to the > respective IPs for testing, but it slowly dawned on me that I could > hook everything up on the lan, to the same subnet and then just reset > the GateWay target on individual machines as needed, for any of 6 > machines. > > So currently I have two internet outlets and two gateway routers on > 192.168.0.0/24 Good job! Many wouldn't have caught that possibility, I bet. > Here's the technical part: > Assume I have loaded a web page that downloads a video to my cache as > it plays. Assume further there are several of these to be played one > by one. > > After playing one, if I reset my GW (and I have also rest > /etc/resolv.conf to use that gw address for dns [probably not totally > necessary]). Right; you could use either to resolve. > Followed by /etc/init.d/net.eth0 restart. You may not even need to do that. I think old ips should work OK with a reset router. > Will the browser, which has not been restarted, now use the new > gateway when I run the next link (or for testing, run the same link > again), or will it continue on the same route (which is still > available), that is, will the browser (firefox) continue using the > original GW until the browser itself is restarted? AFAIK the browser is effectively 'stateless'. It shouldn't need to be restarted for it's behavior to reflect changes to route, ifconfig, or resolv.conf. > I know I could track all this with tcpdump but it gets sort of > cumbersome unless you've memorized the necessary commands to filter > output down to something more usable. I usually get so tangled up > with tcpdump I spend more time on it than the project at hand. I > don't use it very frequently so inevitably spend gobs of time at > `man tcpdump' instead of tending to what I started to do. Yeah, it's confusing. I usually use one of a few incantations: # tcpdump -i eth0 port not 22 that dumps packets on interface eth0 that arent to port 22 (which I was using to connect to the server, and gets messy real fast, as tcpdump itself will be sending over port 22, hence a never-ending cycle of tcpdump reporting its own traffic) or perhaps something like: # tcpdump -i eth0 port 80 that dumps all connections with port 80 on either side. Finally, for you, something like # tcpdump will dump everything. You can then ctrl-C and look through the output (shift-pageUp/Dn if you have scrollback buffer in the kernel) > Why I ask is that the site I'm doing this on requires me to login and > then relocate the stuff I want to see if I have to restart the > browser. No, I do this stuff all the time (to set up vpn from coffeeshops and the like) and I don't ever restart my browser. > I wanted to try to gauge if there was much of a noticeable difference > with the two IP connections. And it would be handy to just step > through the links changine the GW intermittently. Yes, you can do that, but if you put a linux box between the gateways and the network you can use both at once. -- gentoo-user@lists.gentoo.org mailing list
