[Gllug] IPv6 allocation options

[Daniel P. Berrange]

On Tue, Jan 18, 2011 at 12:24:13PM +0000, Mark Brier wrote:
> On Tue, 18 Jan 2011 12:08:02 +0000
> John Hearns <hearnsj at googlemail.com> wrote:
> 
> > On 18 January 2011 11:36, Daniel P. Berrange <dan at berrange.com> wrote:
> > >>
> > > More seriously though the ever increasing use of NAT is bad because
> > > it strongly favours a client-server model and makes peer-2-peer
> > > harder/less reliable. This is not good for resilience against
> > > government and corporate censorship. Increasingly centralized
> > > hosting, under the banner "cloud computing", is being pushed as the
> > > future of highly flexible & reliable web hosting / publishing for
> > > everyone. The wikileaks / Amazon incident demonstrates this is a
> > > exceedingly fragile house of cards as far as freedom from political
> > > or corporate decisions is concerned.
> > 
> > You are a COMMIE PINKO.
> > Hell I bet you use leftie free software which is a cancer that
> > attaches itself in an intellectual property sense to everything it
> > touches.
> > Heck don't you KNOW that the US Air Force has blocked access to your
> > pinko liberal Guardian as it is a threat to the security of the free
> > world?Sir,
> 
> Can we keep it on-topic ... please? Jeez man.
> 
> Your ISP is always going to be the provider of your access, regardless
> of whether they give you a NATted address or a public one. I personally
> don't see how a move to NAT over public or cloud over non-cloud has
> implications for 'resilience against govt/corporate censorship'. Surely
> if govts and companies want to censor the internet, they'll just
> jump into bed with the ISPs?
> 
> Can you please expand and explain for me (and others who are missing
> something here) ... how the .... "wikileaks / Amazon incident
> demonstrates this is a exceedingly fragile house of cards as far as
> freedom from political or corporate decisions is concerned" and how
> this has anything to do with ipv6 or using NAT over public addresses.

In the traditional hosting world, no matter which ISP you choose for your
servers, you basically get the same overall service. If you have problems
with your ISP, you can trivially move to any other ISP by simply relocating
your servers or services. Cloud hosting is appealing to to many because of
the flexibility of billing& scaling up services as traffic demands. If you
just use your cloud hosting company in the same way as your traditional
ISP as a simple OS/machine provider, then you can still trivially move
between providers. This is only the first phase of cloud hosting. Looking
at Amazon who are leading the market in many respects, you'll see that
the next phase of cloud hosting is to actually design & build your
infrastructure around extra cloud services. Amazon are encouraging customers
to make use of things like EC2, Elastic Block Store, Simple Queue Service,
and much more. Other cloud providers will no doubt follow because these
extra services do make it easier to build scalable web services. The
trouble is that all of them are non-standard, proprietry services, so if
you build scalable web services on the cloud with this technology your
IT infrastructure is now locked into the cloud infrastructure of your
Cloud provider and thus cannot freely move to other ISPs. You have very
redundant hardware and software services, but no organization/provider
redundancy as you do with traditional ISP hosting usage.

The point I was trying to make wrt to IPv6 vs IPv4+NAT, is that NAT is
a hurdle to peer2peer applications. IPv4+NAT does not preclude p2p usage,
but it means that apps that want to be robust in the face of multiple
layers of NAT, have to do some quite complex tricks to punch holes through
the NAT(s). Punching holes in NAT also requires that both parties have
access to a common rendevous server for initial connection setup, which
is another failure risk. You could see the effect of this with the trouble
Skype recently had bringing their p2p mesh network back online after its
outage. The large number of clients needing to establish initial connection
caused serious load on the supernodes.  If all clients had IPv6 public
addresses, there would be potentially much less reliance on shared
rendevous nodes to do NAT hole punching, and thus an overall more 
reliable & scalable p2p protocol.

Regards,
Daniel
-- 
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 196 bytes
Desc: Digital signature
URL: <http://mailman.lug.org.uk/pipermail/gllug/attachments/20110118/37e8e501/attachment.pgp>
-------------- next part --------------
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug