Hi all, I am following closely what is written here. Also I am thinking about various solutions to the problem assuming LL uniqueness is not guaranteed (someone told me it is guaranteed for Etherenet is that correct?). Lets say that as an implementers of a File Server we are implementing the extra step of zones. This means basically that we add a zone to each LL source address to identify from which interface it was approaching... Consider the following scenario: Our multihome server is connected through a switch to multiple clients, all on the same broadcast domain (link-local). Lets say now that all the clients are using Link Local addresses. Both Interfaces will view the same Link Local set of addresses but with different zones (different interface...). The routing table/applications will consider them as different entities! I have no doubts that this will create all sorts of bugs in applications that will treat those as two separate entities.
What this means is that now the multihome host will have to apply some zone discovery methods (I don't know how, god help us...). First Question -------------- Any Ideas How The above can be resolved? We are supplying our customers with ability to restrict access from some clients and grant an access to others. Lets say that in a Link Local environment the customer assigns global addresses for all the clients and the server. Since the it is a link Local environment the client contacts the server using the LL address (Is that true?) and all these restrictions are bypassed... Second Question ---------------- Will a client will use the global assigned address to communicate in this case or the LL? Third Question --------------- Lets say, we can't make any assumption about the address (meaning the client may choose either global or LL). Let say we have 3 interfaces of the server connected through a switch to all the clients. Will that mean that the poor network manager will have to assign this restrictions 3 times for each LL + zone on each interface? Comment -------- What a mess... Am I missing something? Shuki -----Original Message----- From: Robert Elz [mailto:kre@;munnari.OZ.AU] Sent: Wednesday, October 23, 2002 7:23 AM To: sasson, shuki Cc: Keith Moore; Brian Zill; [EMAIL PROTECTED] Subject: Re: Link Local Address usage for multi-home host. Date: Tue, 22 Oct 2002 08:47:54 -0400 From: "sasson, shuki" <[EMAIL PROTECTED]> Message-ID: <33CE6457C7003A478381BCD0B584DEC55EAE2C@srmoon> | Link Local addresses should be globally unique addresses so scoping | shouldn't be a problem (64 bits should do...). It is all evry nice to desire this, but useless unless we have some mechanism for actually making it happen. And we don't. Take SL addresses for a minute (where the issues this way are a little easier - LL's need to be available to a node before it actually starts communicating on the net, which makes them very hard to deal with). SL's have a whole bunch of bits that was decided at Yokohama, and not clallenged on the list, will be available to be considered as part of the SL address - easily enough for everyone to have a a unique SL address prefix (not routable, but unique). But actually making that happen would require a whole identifier assignment bureaucracy, and no-one wants to actually create (yet another) entity like that. So, instead, sites just invent their own identifiers (perhaps in co-ordination with other sites that they might some day want to exchange SL info with) and they're not expected to be unique. kre -------------------------------------------------------------------- IETF IPng Working Group Mailing List IPng Home Page: http://playground.sun.com/ipng FTP archive: ftp://playground.sun.com/pub/ipng Direct all administrative requests to [EMAIL PROTECTED] --------------------------------------------------------------------
