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

[Robert Raszuk <[e-mail]>]

Hi Sander,

There is nothing there specific to wide communities.

Transitive across AS is applicable to all TLV encoded communities.

We may add a new flag to mark transitiveness across same operator ASes.

We agreed to dropped C flag as per my comment so forget it.

Local vs registered I already explained .. this is not only about wide
comms. This is also about geo encoding. This is about meaning in the
community (even large) which is understood without explicit policy code to
handle it.

The comment below needs to edited if it goes to standalone doc - you are
right. I just copied for now from main spec.

Type, length size is recommended to be of two octet each.

Context-AS is agreed already among everyone.

Cheers,
R.


On Sat, Sep 10, 2016 at 12:40 AM, Sander Steffann <[e-mail]>
wrote:

> Hi,
>
>    +------+-------+----------------------------------------------------+
>    | Bit  | Value | Meaning                                            |
>    +------+-------+----------------------------------------------------+
>    |  0   |   0   | Transitive across AS boundary                      |
>    |      |   1   | Not Transitive across AS boundary                  |
>    |  1   |   0   | Transitive across confederation boundaries         |
>    |      |   1   | Not transitive across confederation boundaries     |
>    |  2   |   0   | Local community value.                             |
>    |      |   1   | IANA Registered community type or value.           |
>    | 3..7 |   -   | RESERVED - MUST be zero when sent and ignored upon |
>    |      |       | receipt.                                           |
>    +------+-------+----------------------------------------------------+
>
>
> I may be missing something here, but how is this going to work with
> routers that don't know about wide communities? Aren't communities marked
> as not transitive across boundaries going to leak across anyway and cause
> unexpected problems there?
>
>    Flags are defined globally, to apply to all wide community container
>                                   types.
>
>
> This common header feels overengineered to me. How about just type and
> length, no magic?
>
> Cheers,
> Sander
>
>