Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!think.com!zaphod.mps.ohio-state.edu!swrinde!cs.utexas.edu!smoot From: smoot@cs.utexas.edu (Smoot Carl-Mitchell) Newsgroups: comp.protocols.tcp-ip Subject: Re: Proxy ARP question Message-ID: <20216@cs.utexas.edu> Date: 27 May 91 03:10:28 GMT References: <97879@lll-winken.LLNL.GOV> Organization: Dept of Computer Sciences, UTexas, Austin Lines: 28 In article hedrick@athos.rutgers.edu (Charles Hedrick) writes: >Proxy ARP seems to have been invented originally simply to handle >the problem of systems that did not implement subnets. I believe >the first implementations worked only with addresses on other subnets >of the same network. By definition Proxy ARP can only respond to hosts which do not understand subnets. If the machine sending the ARP request understood subnets, it would use the usual IP routing mechanism to route its packets. Proxy ARP is a surrogate for real routing. The host which does not understand subnets thinks the whole network (which may be subnetted) is directly attached to its own network interface. We did add a "feature" which lets Proxy ARP check the routing table of the responding router. This is useful in circumstances where a particular network segment is not really subnetted. This allows you to establish the subnet mask simply on the router an leave all the host network masks as if they were on an unsubnetted network. All that is required is to establish static routes to each of the ``pseudo'' subnets accessible via that interface. This is a useful feature for sites which are slowly migrating to a fully subnetted network, but either have machines which do not understand subnets or are forced to slowly migrate because of time contraints. Proxy ARP is useful, but using true subnets is a much better alternative. -- Smoot Carl-Mitchell, Texas Internet Consulting smoot@tic.com, smoot@cs.utexas.edu