Charlie,

"Charles E. Perkins" wrote:
> 
> Hello Brian,
> 
> Your points are very well taken.  My main suggestion for the
> immediate term was to relax some restrictions on well-known
> anycast addresses located in their subnets.  It is really
> easier if there's just one host, too.  The other things
> will take more work.
> 
> >From that context, I'll just add a few more comments inline:
> 
> Brian Haberman wrote:
> >      1. Host-to-router notification protocol (this is taken care of by
> >         changes to mld proposed in draft-haberman-ipngwg-host-anycast)
> 
> What happens if you have a host on a link responding to an anycast
> address on that link?  Then other hosts on that link won't be able
> to easily discern whether the router "liked" the anycast addressable
> host or not.  I can imagine some very nasty ways to avoid even this
> scenario, but I can't believe we'd really want to enforce them.

I haven't thought it all through, but perhaps there is a way to
handle the on-link case with ND.

> 
> >      2. Security: at a minimum some form of authentication to allow
> >         routers to determine if hosts are allowed to join an anycast
> >         group
> 
> Similar comment to the on-link case for (1).  Furthermore, the
> security requirements for an anycast group depend on the nature
> of the anycast group, don't they?  It might typically be quite
> unimportant to securely restrict participation in an anycast
> group geared towards local broadcast of streaming audio of classical
> guitar music, or a webcam of the cows in the pasture.

Definitely, the nature of the group will help determine the security
requirements of that group.  However, I would not advocate the use of
anycast for some of your examples. :)

The security requirements of membership in an anycast group, I believe,
is much more important than multicast security requirements.  The
primary reason is that a rogue node can launch a very effective DoS
attack by joining an anycast group.  Since the packet goes to one
member of the group, a rogue can easily blackhole traffic to the
group for portions of a network.

> 
> >      3. An Anycast Architecture doc that pulls all the pieces together
> >         and concretely describes how pieces interact, the pros and cons
> >         of anycast usage for intra-domain and inter-domain
> 
> Right!  But the latter is often true for any service.  The anycast nature
> of the addressability of inter-domain isn't appreciably harder, otherwise,
> than handling anycast groups in general, is it?

It is harder in the fact that the current mechanism of anycast routing
requires host routes.  In the intra-domain case, anycast routes may
still aggregate, but that may not be true inter-domain.  Especially
when members of the same anycast group span multiple, independent
networks.

> 
> >      4. Possibly a draft that documents any impacts on any existing
> >         protocols (routing protocols, TCP, etc.)
> 
> This would be very important.
> 
> > It should also be noted that this is probably way too much work to do
> > in the IPv6 WG.
> 
> It's way too much work for me too :-)  I'm not really suggesting that
> the general case be made completely open prior to this work getting
> done.  I was more suggesting that there are useful special cases that
> do not present any appreciable downside.

Regards,
Brian
--------------------------------------------------------------------
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]
--------------------------------------------------------------------

Reply via email to