Let me make some comments fundamentally for background. It's increasingly considered useful to minimize the amount of configuration that an end station needs to do before becoming active. Servers and routers are special cases. DHCP is stateful: the DHCP server remembers what addresses have been assigned. Apple and Microsoft alternatives are dynamic stateless alternatives. They select a tentative name, and then broadcast sequentially selected addresses until they find one with no conflict. BTW, this is similar to the way NetBEUI checks for name uniqueness. CLNS and IPv6 alternatives listen for a high-order prefix defining alink-local or site-local part, and prefix these part(s) in front of MAC addresses. Still mostly stateless, but more cooperative. >At 10:37 PM 1/6/01, John Nemeth wrote: >>On May 29, 5:24am, Craig Columbus wrote: >>} >>} OK. I can accept that Microsoft (or Apple for that matter) would do >>} something like this and then expect the world to revolve around >> >> Actually, as Howard mentioned, neither of these companies >>initiated the protocol > >It can be argued that Apple initiated the particular protocol that we have >been discussing, that is, the Microsoft Automatic Private IP Addressing >method. The client sends a gratuitous ARP 10 times broadcasting the >network-layer address that it wants to use. If the address is in use, the >client selects another address. The creators of AppleTalk, including >Gursharan Sidhu, Ron Hochsprung, and Alan Oppenheimer own a patent that >reads essentially just like that. > >The patent is from 1984. At that time IP networks were managed by computer >scientists. Apple had the brilliant idea that ordinary people could set up >and manage networks. I think we should give credit where credit is due. As >Chuck mentioned, in the 1990s Microsoft also tried to make file and print >sharing easy, but the majority of the credit should go to Apple. > >Apple could have imposed the dynamic network-layer addressing patent on the >industry but perhaps it was too specific. (It has a bunch of LocalTalk >specifics in it.) Also, they probably let it go because they recognize the >value of furthering the ease of use of IP networks. People who will be >connecting their home appliances together don't want to understand IP >addressing, subnet masks, etc.! And how about ad hoc networks in training >classes, on long plane rides, in hotel lobbies, on the beach, etc. &;-) > >John makes some other very good points below. I don't want to detract from >them, but I just had to make the point again about AppleTalk. It's unfair >to not do so. > >Priscilla > > >>} them. However, I'm confused as to the benefit. Why would anyone want a >>} non-assigned default IP address to appear on their network? Do they really >>} think that people will implement a non-RFC1918 compliant address space just >>} to save configuration time? (Actually, I can think of several cases where >> >> It does save configuration time, since this is for cases where no >>configuration at all happens, most likely due to the lack of a real >>administrator. >> >>} How do Internet backbone routers (BGP ASs) deal with this traffic? >> >> They don't. There is a reason why this address range is called >>"link local". It's only useful within a single network segment that >>isn't connected to any other networks. >> >>} Let's say that I want to take the easy way out and I connect a small >>} network to the Internet via an ISP. I'm not running NAT, but I'm running >>} the 169.254 addresses inside my network. If I've got a static route to an >> >> Then, you're SOL. To connect to the Internet, some kind of >>configuration must happen (even, if it is just a box running NAT on the >>outside interface and a DHCP server on the inside interface). >> >>} ISP public address, and we're not exchanging routing information, I can't >>} see how this traffic would ever get back to my network. If I'm exchanging >> >> It wouldn't. >> >>} routes with an ISP (via BGP or some other interior protocol), where and how > >} do the 169.254 routes get filtered? There has to be some mechanism, or >> >> It should be filtered at the network ingress point. >> >>} there would be thousands of summary routes back to 169.254 showing up on >>} the Internet table. >> >> 169.254 should never ever show up on the Internet, although I >>wouldn't be surprised if it did. I've seen some pretty large ISP's put >>RFC-1918 addresses on the global Internet, which is also a no-no. >> >>} Any help in understanding this is appreciated. >> >> The purpose of this is to setup small impromptu isolated networks >>which often don't have an administrator with no configuration at all >>required. > > >________________________ > >Priscilla Oppenheimer >http://www.priscilla.com > >_________________________________ >FAQ, list archives, and subscription info: >http://www.groupstudy.com/list/cisco.html >Report misconduct and Nondisclosure violations to [EMAIL PROTECTED] _________________________________ FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED] _________________________________ FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
