Re: RPSLng task force outline
- Date: 05 Oct 2001 08:22:29 -0700
On Fri, 2001-10-05 at 04:19, Mark Prior wrote:
> At 4:23 AM -0400 5/10/01, Florent Parent wrote:
> >Please, take a look at the following task force TODO list and send
> >comments/additions to the list. We need to properly define the scope of
> >our work.
> >
> I think that if we are going to open Pandora's Box and start fiddling
> with RPSL then I would like us to consider revisiting what I consider
> is a defect in RFC 2622!
>
> There isn't currently an iterator in RPSL and I think it would be
> helpful if there was one. Maybe Cengiz can explain why it was
> rejected last time but I will illustrate how I would use it if there
> was one :-)
This was discussed. Some people did not like procedural constructs.
>
> While working for Connect I created a large number of AS set objects
> that listed our customer's AS, their customers, etc. If I was a peer
> of Connect and wanted to filter them via AS paths (or just wanted to
> get my policy correct :-) I would like to use these sets rather than
> reinvent them but there is no easy way to do that.
>
> Connect (AS2764) has an AS set called AS2764:AS-TRANSIT that includes
> a list of ASes that provide transit to other ASes and then for each
> AS in that set there is another set called AS2764:AS-CUSTOMERS:ASx
> (where ASx is that member AS).
>
> As a peer I could use these AS2764:AS-CUSTOMERS:ASx objects but there
> is no mechanism to use AS2764:AS-TRANSIT to iterate through them so
> my policy would need to explicitly name them all and somehow keep
> them up to date. What would be better is something like
>
> import: from AS2764
> accept interate over AS2764:AS-TRANSIT for
> <^AS2764+ TheAS+ AS2764:AS-CUSTOMERS:TheAS*$>
>
You probably want to make this like perl so that operators are familiar
with it.
foreach x in <any rpsl set> { filter expression }
I suppose the semantics would be OR over the loops expansion.
You also probably want a structured policy version of this
foreach x in < any rpsl set > {
from AS2764 accept x
}
> Mark.
> --
>
|