RE: Question on scopes involving IPv6 addresses

[Jim . Bound]

Having the scope be part of the IPv6 address or having other distinguishing attribute in the NLA (which is null now) was discussed and rejected. I was supportive of this idea.   But it does add an entire address space management part to IPv6 site local addresses.  My input is don't go there. Implementations will have to have scoping code following the scope-arch-02 draft architecture.  How that is done is and should be implementation specific.  I think maintaining tables in an implementation is a question of the type of implementation.  I think its better to extend existing OS kernel protocol control block structures and search algorithms to be scope aware.  This also means that information can be extracted by user space applications for management of scopes and for source/destination address selection.  But all should use the assumption that FF05 prefixes will only exist within a site and will not overlap sites, because they can be duplicated.   /jim       Going through scoped address architecture it provides a mechanism to distinguing the same address belonging to different scopes. What I am wondering is that   a) Couldnt the scope index have been part of the IPv6 address. b) For upper layers like UDP which identifes a connection based on addresses wouldnt it be better to put the scope index into the address itself.   One scenario is that we could have the same [source src port destination dst port ]  pairs from different scopes as in the scenario earlier seen. Is the upper layer suppose to handle this in a different manner like having scoped tables.   Thanks Alex