Summary of small meeting with merit of the last ripe-81++ issues

  > Jessica Yu <jyy at merit.edu> writes:
  > 
  > 'Descriptions of interas policies do  not  replace  the  global  policy
  > described  in as-in, as-out and other policy attributes which should
  > be specified too.  If the global policy mentions  more  routes  than
  > the local policy then local preferences for these routes are assumed
  > to be equal for all links.  If a route is only  referenced  in  some
  > interas-in/out  attributes  and  not  in  others  it  is assumed not
  > announced/accepted on the links concerned (see the example above).'
  > 
  > **(Replace the above paragraph with the paragraph below)
  > 
  > ** When both as-in,as-out and interas-in,interas-out are used, for
  > the routes mentioned in both set of attributes, the preference  
  > defined in interas-in and interas-out will take precedence for the particul
  > ar
  > interas connection point identified by <local-rid> and <remote-rid>.  For 
  > the routes which are not mentioned in interas-in and interas-out, their
  > preference will be using what defined in as-in.  If a route is only
  > referenced  in  some interas-in/out  attributes  and not in others it is
  > assumed not announced/accepted on the connection concerned. 

  > **The key difference between  interas-in/interas-out  and  as-in/as-in
  > attributes  is  the former describes a more specific inter-AS policy 
  > based on multiple connections between ASs and the latter the general
  > inter-AS policy.  The  general policy should
  > always  be  defined. The more specific inter-AS policy
  > should only be defined when such a  policy  really  exists  and  the
  > implications of setting such policies are fully understood.

Sorry but this is far less clear. It assumes that the preferences
on the as-in and as-out attributes are comparable to those in the
as-in and as-out attributes which we agreed they are not.
It also implies that you can specify some part of the global
policy in as-* and some other part of the *global* policy in
interas-*.

Allowing intermixing of the two is going to be disastrous both for clarity
and maintainability. While I admit that disallowing it will lead to some 
repetition in the definitions, having complete global and local data
without combining the two sets is far more important.

Also this will encourage proper splitting of ASes if "local" policies
get too complicated. This is exactly what we should encourage.
See recent discussion on CIDRD.

This is the n-th time where we clash on the *fundamental* difference in
opinion that interas-* is a *local refinement* of policy between two
ASes whereas the global inter-as policy is described in as-*.
It seems that the particular style of discourse we have had about this
has not brought this into the open properly.

In order to bring this out I propose to leave the paragraph as before and
add the: "Any route spcified in interas-in/out but not
specified in as-in/out is assumed not accepted/announced between the
ASes concerned. Diagnostic tools should flag this inconsistency as an error."

Daniel



-------- Logged at  Mon Aug 22 17:38:24 MET DST 1994  ---------