[routing-wg]Draft minutes routing-wg RIPE 51

Dear all,

below are the minutes provided by Rene Wilhelm of the RIPE NCC. My  
apologies to Rene for not forwarding this earlier to the list as he  
was quite quick in producing them.

Joao Damas
ISC

==================================
Draft Minutes Routing WG at RIPE51
==================================

Thursday 11 October 2005, 14:00 - 15:30
Chair: Joao Damas
Scribe: Rene Wilhelm

A. Preliminaries (Joao Damas)
-----------------------------

- Minutes from RIPE50 routing-wg

   These were approved, are now final


- Chairing of the group

   Joao explained the chair and co-chair of the group agreed to  
switch roles:
   because Joachim Schmitz often cannot make it to RIPE meetings,  
they felt
   the group was better chaired by Joao, with Joachim providing input in
   a role of co-chair. For the same reason, Joao now asks for a second
   co-chair, someone who could chair the sessions at a RIPE meeting in
   case Joao cannot attend (e.g. due to illness).  Those interested  
should
   talk to Joao; the new co-chair will be appointed at the next ripe  
meeting.


B.  Discussion on on route flap damp(en)ing - Philip Smith
----------------------------------------------------------

http://www.ripe.net/ripe/meeting/ripe-51/presenations/pdf/ripe51-bgp- 
flap-damping.pdf

Regularly people ask about the status of RIPE 229: is there and update?
are the recommendations still valid?  At the same time others contact  
the
authors stating the document has lost its validity, route flap damping
not being useful, counterproductive.

Philip presented the issue at RIPE50 in Stockholm, but no conclusion
was reached in the discussion that followed. So the discussion
continues here, with more people attending the meeting in Amsterdam.

- Should RIPE229 be declared obsolete? should it be modified?
- Is flap damping bad for your network?
- Is it needed at the Internet edge ?
    (i.e. ISPs not providing transit to any other AS)
- Is it needed in the Internet core?

Questions/comments:

Geoff Huston: we did flap damping for the wrong reasons. we thought
edges were flapping for reasons not related to the BGP protocol, but
now we find the protocol is one of the main reasons for routes to
flap in the Internet. Flap damping doesn't work. Option 1, obsolete,
is what most people do.

Flap damping is actual harmful. not only causes you problems but  
elsewhere too.

If you reopen the work, ask why major vendors have different defaults.

Patrick Gilmore seconds Geoff, option 1 is much much better.
 From what he has seen in the past flamp damping is more harmful,
routers with damping enabled died first in times of crisis.
Is there any reason whatsover not to declare it obsolete?

Philip: ISPs at the edge use it for stability

Geoff: If you have a prefix from two paths and one path is flapping
why wouldn't the other do the same? generally, dynamic noise
in updates depends number of peers and granularity; if you have
multiple path same prefix, would it not be a rare occurance for
one to be stable and the other not? More likely, both flap at same time.

Kurt Kaiser: combine flap damping with max prefix?

Joao notes that no one is speaking up for keeping it on as a work item.
However, he feels that just declaring RIPE 229 obsolete is not enough;
it should be accompanied by a small document which explains why, in  
general,
flap damping should no longer be applied and describes the  
exceptions, if any,
where flap damping might still be useful.  Geoff Huston and Philip Smith
agree to look into this.


C. RIS update. Arife Vural, RIPE NCC.
-------------------------------------

http://www.ripe.net/ripe/meetings/ripe-51/presentations/pdf/ripe51- 
ris-update.pdf

Arife presented an update of the status of the Routing Information  
Service.

Interest in raw data is increasing, with a stable number of unique  
visitors
Looking Glass, ASinuse and BGPlay are most popular utitilies.

The IPv6 version of RISreport is now online
(http://www.ris.ripe.net/risreport/ipv6/)

Worked on improving RIS performance, changes to RIS infrastructure
are being prepared and will come online in a couple of weeks.
This will provide faster database replication, less time lag between
receiving BGP update and seeing it in the database. Also will allow
to have more data in the RIS DB, upto 1 year.

Joao: You are adding more and more collectors. How many are enough?

Arife: we have some offers, idea to put one in South America.

Henk Uijterwaal: we were appraoched at last RIPE meeting by LACNIC,
this week heard they have equipment, we can install when we like.

Shane Kerr: we are not sure when to say we have enough; plan to come
up with a measure to see how well covered. Not sure what users
need, time lag, history. We know people get raw data but don't
know what people do with it. If you use the RIS, come talk to us,
we'd love to hear what you're doing.


D.  Route Aggregation Best Practice -  Mike Hughes
---------------------------------------------------

Action as result of what happened in the LINX50 meeting in August

LINX used to have a rule regarding aggregation, "members shall  
aggregate"
but this rule was never actively enforced.  LINX council created a draft
route aggregation policy which was discussed at LINX50.

Member response: LINX must not interfere in routing, not adopt and  
enforce
a policy for route aggregation, but we do want to the right thing.
New rule: "members are encouraged to aggregate"   It was suggested
to bring the draft policy to the RIPE routing-wg, where perhaps it
could be converted into a BCP document.

Opinions?

Gert Doering: I see lots of deaggregated garbage from peers; when asked
they reply "this is coming from my customer, can't control it",
but they could if they wanted to. Might be nice to have a BCP,
but unless there's some force it will not have any effect whatsoever.
No peer or depeer pressure that will work. depeering only felt
if done by a big provider, but those have other reasons than
technical to peer/depeer.

Patrick Gilmore: I'm also pessimistic, but ISP to ISP could work,
a BCP might help there.

Ruediger Volk: nothing new to add, trying to convince people to
do the right thing without a paper is not going to help. A BCP
will help a bit, but is no guarantee.

Geoff Huston: since 2002, the fraction of more specific prefixes in
the routing table is stable around 50%, but the amount of addresses
is growing from 15% to 18%.

Mike Hughes: people already said there must be some place to start

Joao will evaluate the feasibility of a BCP on route aggregation
with some volunteers, including Mike Hughes and Philip Smith.


Z. AOB
------

no other business.