Stephane,

> On Jul 20, 2015, at 2:21 PM, <[email protected]> 
> <[email protected]> wrote:
> 
> [SLI] Do we really need to differentiate from a policy point of view ? from 
> an import policy perspective, matching a tag, means learning the tag value 
> available in the protocol (if available) and when the route ins inserted into 
> RIB the tag value is copied from the protocol value if not overrided by 
> import policy action; from an export policy perspective (talking about export 
> from rib to protocol), matching a tag means matching the tag value in the RIB 
> (which may come from protocol or not),  setting a tag means fill the protocol 
> field if available. From a RIB point of view, the tag associated with the 
> route is protocol agnostic, even if the protocol does not support tags in 
> encoding you may associate a local tag for policy processing.
>  
> Having two types of tags is also possible : protocol-tag and local-tag but I 
> see more complexity and do not see more flexibility : but maybe there is some 
> use case that I do not see.

The messy detail with this attribute is that while it's useful as a generic 
policy element, in specific protocol context it needs to have differing 
constraints.  OSPF has one set of constraints, RIP a slightly different one 
(zero is reserved), and ISIS has different sizes with some option for one or 
two tags plus the 64-bit tag previously discussed.

This set of context specific constraints probably removes some level of the 
flexibility that you'd want for it to be a generic marker - unless you can live 
within the least common denominator.

-- Jeff

_______________________________________________
rtgwg mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/rtgwg

Reply via email to