Re: DSTM Considered Harmful [Re: (ngtrans) final ngtransagendaforIETF-53 in Minneapolis]

From: Octavio Medina (octavio.medina_at_no.spam)
Date: Tue Mar 19 2002 - 11:10:55 PST


Christian Huitema wrote:
>
> > On a deployment viewpoint, a MS implementation would be very useful,
> > given we have good v6 app support in Linux and BSD. Can any of the
> MS
> > people comment on their own plans here?
>
> I am personally not convinced of the urgency of implementing and
> deploying DSTM. The stated goal of DSTM is to let users of an IPv6 only
> network gain IPv4 connectivity. Think about it: IPv6 only networks. Does
> not look like a year 2002 problem to me. I would rather concentrate on
> making sure that users in an IPv4 only network can get some IPv6
> connectivity.

If we can offer a reliable method to assure IPv4 connectivy inside an
IPv6 only domain, people would have the choice between setting up IPv4
private addresses + NAT (and all the associated problems) or IPv6 +
DSTM. People are using private addresses + NAT today, in 2002.

>
> I am also not convinced that DSTM is the right solution to the stated
> problem. It goes to great length to assure that the IPv6 only nodes can
> get a global IPv4 address, but it is very unclear that we will have
> enough available IPv4 addresses to allow that. DSTM attempts to reduce
> IPv4 address consumption by "time sharing" the address, but we have
> experience and theory that tells us that time sharing does not work
> well: in short, all nodes tend to need addresses at the same time.
>

That's what the ports option is all about, to offer that port sharing
possibility (with its limits) and reduce the number of required
addresses.

> If you remove the "global address" property, then DSTM is in fact an
> inferior solution, compared to NAT-PT (provided we remove the DNS ALG
> from NAT-PT.) At least, with NAT-PT, I don't even need an IPv4 stack in
> the IPv6 only nodes, and I also don't need any address allocation
> protocol between the v6 only node and the server.

In NAT-PT you assume you have IPv6-ready applications, capable to send
IPv6 packets. That is not the case for most applications (specially for
Microsoft's). For NAT-PT to be useful, we would need a big number of
applications becoming IPv6 aware, which is not the case in 2002.

-Octavio



This archive was generated by hypermail 2.1.7 : Fri Oct 06 2006 - 00:00:31 PDT