RE: Auto-tunnel Rant

["Dave Thaler" <dthaler@Exchange.Microsoft.com>]

> > On Fri, Mar 23, 2001 at 07:32:48AM -0800, Kevin C. Almeroth wrote:
> > > Well, I'll add my voice to the rest of the noise.  As I see it,
> > > auto-tunneling is fine.  Why?
> > >
> > > 1.  Let's not get in the business of protecting people from
> > >     themselves (in reference to Dino's comment about large
> > >     fanout is bad...  it is, sure, but who cares?)
> >
> >         Those of us who are customers of router vendors (not
> > server/software vendors) have concerns about this impacting our
> > routers should they spend time creating tunnels.  large fanout
> > is bad in this case as to get upgrades to the cpu/power in these
> > takes 18 months as compared to the server market that gets a faster
> > cpu every few months and it is easier to add extra machines 
> when it's
> > needed.
> 
> That is the idea behind the CastGate proposal, it even allows 
> to distribute the
> tunnel end-points over your network....the CastGate client 
> will use - if
> allowed by your policy -  the Tunnel server reacheable via 
> the least congested
> path!
> I don't need routers in the proposal, I use PC's who are 
> really cheap as
> compared to even basic routers. ;-)

Actually I think that PCs would do for any of the auto-tunnel 
proposals.  None of them (to my knowledge) from this IETF assumed
that replication would be done by a normal router.  Rather,
my understanding is that they all assume a tunnel endpoint server,
where that server could be a PC or could be a special-purpose router
box that is decided for this purpose.

-Dave