Nick,
On Wed, Oct 23, 2013 at 03:45:52PM +0800, Nick Hilliard wrote:
> My thoughts on the draft is that per-se there is nothing wrong with it, and
> it will provide a backwards-compatible mechanism for an interesting
> attribute which is relevant to ixp route server connectivity. The rationale
> for using a new class attribute is that it's not possible to shoe-horn the
> required semantics into mp-{import,export}.
Agreed.
For historical reference, the RSng route-servers provided this functionality
at the inet-rtr level rather than aut-num. (Attributes rs-in and rs-out)
The aut-num level as specified in this document is a bit more general and
centralizes the functionality a bit better than inet-rtr. However, I don't
have a strong opinion whether the updates in this draft are in inet-rtr vs.
aut-num.
Carrying import-via/export-via in inet-rtr does have the possible advantage
that the sieving mp-peering-1 in the draft simply drops out and is a side
effect of the specified inet-rtr. However, exchange points often have more
than one router and this means that policy is duplicated.
> And this is the problem: RPSL is broken enough that it's not possible to
> extend the language in a backwards compatible way, and the only option for
> stuff like this is to create a completely new class for an incremental
> feature upgrade. Although I have no particular objection to this draft, I
> think it shows that RPSL is too broken to put much serious development into
> it, and that it would be more productive to look at creating a new
> alternative.
Firstly, the I2RS folk would probably like to talk to you about this. :-)
Secondly, the amount of work that went into even getting as far as RPSL from
RIPE-181 was daunting. I'd think very hard before considering such work.
-- Jeff
_______________________________________________
GROW mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/grow
