Re: processing order of reach/unreach in rfc2858bis

[Enke Chen <[e-mail]>]

Alex,

> Date: Wed, 20 Feb 2002 10:20:49 -0800
> From: Alex Zinin <[e-mail]>
> Message-ID: <[e-mail]>
> To: Yakov Rekhter <[e-mail]>
> Cc: [e-mail]
> Subject: Re: processing order of reach/unreach in rfc2858bis
> In-Reply-To: <[e-mail]>
> References: <[e-mail]>
> 
> Yakov,
> 
> > no, it was an unintentional omission.
> 
>  Then I'm thinking that Option 1 could be made better if
>  modified to say that the prefix should be processed as if
>  it only appeared in the highest priority field, where
>  the field priorities are defined as (highest left):
>  
>    NLRI, MP_REACH_NLRI, {MP_UNREACH_NLRI or WITHDRAWN}
> 
>  The difference is that we process a prefix as if it was
>  in one field only (perform only one operation) instead
>  of processing it in all fields (and potentially all
>  instances of the prefix) in a specific order,

For folks who care about (or need to deal with) code deployment and
compatibility, please note that the majority of the depoyed routers
process the prefixes in the sequential order. To do it differently,
we will need a strong reason why the sequential processing is not
good enough, and be willing to deal with compatibility and transition
issues.

> which is more consistent with the base spec, where we instruct
> to ignore the prefix instance in WITHDRAWN.

Please note the sequentail processing yields the same result as
ignoring the WITHDRAW prefix.

-- Enke