Re: [Idr] Request to adopt draft-heitz-idr-large-community - Working Group Adoption call (9/6 to 9/20)

[Robert Raszuk <[e-mail]>]

Jared,

Agreed with your last mail.

My conclusion of today's discussion leads me to believe that 4:4:4 has
number of use cases and will allow to use some of current operational
practices (which took million of manhours already and is likely to take
more millions of manhours to adjust 2:2 to 4:4:4 by rewriting current RPL
policies (or similar) and maintain those).

It is obvious just based on very simple example that 4:4:4:4 would simplify
number of policies and avoid to heavily overload last 4 octets in 3x4.

Yet in few more cases there is evidence for the need of 4:4:4:4:1 or :2 and
as indicated 4:4:4:4:16+1 would allow to use IPv6 prefix ... Likewise Jeff
mentioned his customers asking him for RFC5701 to embed text.

The easiest way to address all of the above without creating separate BGP
attribute for each is to encode them into a TLV format where TLV can be in
the front (aka common header) or at the back.

Maybe there are other ways to accommodate all of the above in efficient way
though ...

Best,
R.


On Wed, Sep 21, 2016 at 4:23 PM, Jared Mauch <[e-mail]> wrote:

>
> > On Sep 21, 2016, at 10:04 AM, Robert Raszuk <[e-mail]> wrote:
> >
> >  sometimes worse is better.
> >
> > ​Well if this is an argument for IDR chairs to adopt a draft then I rest
> my case .....​
>
> I think you misread this two ways, namely intent.
>
> Simple > complex
>
> I get the impression you favor a well defined IETF driven structure of how
> networks are architected and organized.  Some people view that as complex
> and imposing.
>
> opaque > well defined
>
> Operators run the networks the way they do for the benefit of their
> business needs.  Why so few well defined communities years after rfc1998
> was published?  I think it’s clear, there is not tooling consensus amongst
> the carriers nor the customer demand to push convergence at this point.  I
> await to be corrected or tapped on the shoulder in this regard, but the
> opaque nature of 1997 and of -large- where we could use that last variable
> as a proxy code point for something like blackhole, prepend, some new
> routing algebra seems feasible.
>
> I’d take a close look at how we can do this in XR RPL, it’s quite good at
> tokenizing ($1,$2) out of 1997 communities and allowing an operator to
> write code around that.  Wish other vendors had such code, would make
> writing policies much easier.
>
> - Jared
>
>
>