[6bone] RFC2772 rewrite -- bigger scope goals

Robert Kiessling Robert.Kiessling@de.easynet.net
17 Nov 2002 12:57:36 +0000


Nicolas DEFFAYET <nicolas.deffayet@ndsoftware.net> writes:

> On Sat, 2002-11-16 at 16:23, Robert Kiessling wrote:
> > Here is a rather radical proposal:
> > 
> >         Peerings between 6bone sites MUST NOT carry any other routes apart 
> >         from 3FFE and a summary route for 2001/3.
> 
> You can have unstability problems with RIR space...

You can have stability problems in IPv4.
Does it matter for this discussion? No.

What matters is what actually happens, i.e. what is *likely* to cause
problems. And here we saw several instances of unmaintained and badly
responsive 6bone site, and sites running two-year old copies of
depricated BGP software.

For example, all of the LIRs I have spoken to understand the problem
with long-distance tunnels and are getting rid of them, while at the
same time a 6bone site proudly announced to have more than 90 tunnels.

> A lot of ISP use their RIR space for do experimental stuff (it's more
> easy for a LIR to get a sTLA than a pTLA for do tests...)

I anticipate filtering of announcementes from such sites in RIR space,
which solves this problem.

> sTLA owner can don't have IPv6 experience because it's not a
> requierement for get a sTLA.

Entities operating RIR allocations will have BGP experience, which is
more relevent. And they have a view and sense of real network
topology.

> I think that your solution is not good and will kill the 6bone.

It might be more productive if you explained concretely why you think
it's a bad idea. I don't see any indication why this would "kill
6bone". What do you mean by this?

> Why limit 6bone space if the pTLA are expected to provide production
> quality service ?

In all practical terms, this requirement from 2772 is ignored. As it
has to be, since "production quality" is too vague to be acted
upon. And as the discussion here shows, there is reluctance to detail
this, e.g. by requiring 24h response time.

> I don't understand how RIR space can get problems with 6bone space...
> Please explain me.

I don't understand your sentence, sorry.

> My router will don't do better announces with RIR space than 6bone
> space....

You don't get RIR space because you are not LIR, don't meet the
requirements to become LIR and most likely wouldn't be prepared to pay
for it.

In other words, you operate it as a hobby, while LIRs operate
professionally. And it's not just you, it's many other 6bone sites
too.

This is not to say that this is a bad thing. In the opposite, much
good came out of private initiative. However, there are global
operational impacts, and they must be limited.

RIRs have the background of production network operation in the IPv4
world. Serious, quality-oriented IPv6 backbone operation is what we
now need to bring IPv6 forward.

Robert