Re: new proposal for rpslng

[Florent Parent <]

--On 2002-01-15 14:03:47 +0100 joao@localhost wrote:

> First of all, sorry for not having commented earlier.
>
> Let me start by showing my astonishment about the fact that of the 4
> suggested methods for expanding RPSL described in 2622, you seem to have
> jumped directly to the last one, the least desirable one.
>
> Apart from the obvious software problems the current proposal may (will)
> generate,
? Clearly, this implementation requires an update to the current tools to 
understand the new syntax.

> I think there is an additional one, which for me is quite
> important: user confusion.
>
> I would rather have a route6 class and import6/export6 attributes (in the
> autnum object). It is cleaner and it ehklps to avoid user mistakes, both
> when generating the data and when reading it.
At the cost of introducing new attributes and class for IPv6 only. I had 
initially considered that "route", but you end up adding many new 
attributes and classes to add IPv6. The route taken here is more like 
making RPSL "IP independent", ie. not adding any new classes or attributes 
that are specifically tied to an address family.

> One other thing I found surprising while reading the current proposal was
> the fact that, while there seems to be discussion about even mpls-vpn in
> RPSL (which I myself believe to be outside the scope of RPSL),
I don't see any such reference in the document.

> there is
> mention of ipv6 and multicat tunnel description. While one should be able
> to describe routing policies for native IPv6 networks, I believe if we
> leave out a way to describe tunnels, 6to4 gateways, etc the use for the
> IPv6 RPSL will be very limited. As a matter of fact, this was the part
> where we asked for more input from people running networks, so that we
> could get a set of clear requirements that the language extensions should
> address, before jumping into an i-d.
Tunnel description was removed from the current document as I needed more 
thoughts/discussion on the subject. I fully agree that we have to work on 
this.

Thanks Joao. I hope more discussions will occur during the conference.

Florent.