On Fri, Jan 04, 2002 at 09:38:55AM -0600, Julian Opificius wrote: > But I'm afraid your answer to my last lingering question was a little > unclear, I think because I didn't phrase the question properly :- > ... > My question should have been:- how does the machine running bind satisfy > DNS requests from OTHER machines on local (private) network if it doesn't > look at hosts ?
No, I see now what you're looking for. I didn't finish the last yard of the run--how do you disseminate information to OTHER machines on the LAN? The resolver routines on the server are only used for queries from programs on that server. This doesn't make provision for answering requests from other machines on the LAN. Now, while it's conceivable that you could do some hack to use the server host services using something like RPC or rsh, why bother? if you want that, there's a solution already (DNS). In a hosts-based network, then, you typically distribute the hosts file to each machine in the LAN. If it's an all Unix/Linux network, you can use yellow pages--er, YP--or, most simplistic, a cron job that shoves or pulls the file on a regular basis from the server. This was the state of affairs, BTW, just before named/DNS was introduced. If there are Windows machines involved, the same host file works (it must be in either C:\WINDOWS for Win9x or C:\WINNT\SYSTEM32\DRIVERS\ETC for WinNT/2K, and I'd have to go look for WinME or XP.) Automated distribution is a bit more problematic with Windows; you either get a third-party 'rsh', or run somehing with the scheduler. In either case, each machine is using its own, local resolver mechanism to access the distributed hosts file. DNS is really quite easy to set up--so easy that, even if you're not on a WAN or connected to the Internet, it's still quite reasonable to create your own internal zone file and run the DNS server on one internal machine to serve all your other boxes. This has the advantage that, if/when you DO tie in with the great outside world, all the changes are on your server--the other machines in the network are already looking to the server for DNS resolution. This configuration reflects how most of the early Internet worked when long-haul connections were via UUCP, but local networks had gotten too unwieldy for hosts-based name resolution. I hope this explains a bit more. If anything else is unclear, please just holler! Cheers, -- Dave Ihnat [EMAIL PROTECTED] _______________________________________________ Redhat-list mailing list [EMAIL PROTECTED] https://listman.redhat.com/mailman/listinfo/redhat-list