Minutes were taken be Curtis Villamizar except the portion of the
meeting where the RPS Charter was discussed. Minutes in this part of
the meeting were taken by Cengiz Alaettinoglu.
Agenda Bashing (Cengiz Alaettinoglu)
RPSL transition (David Kessens)
ARIN (?)
RIPE (?)
C&W (?)
RA (Merit, Craig Labovitz)
ANS (Curtis Villamizar)
ISI (Cengiz Alaettinoglu)
RPSL Changes in New Draft (Cengiz Alaettinoglu)
Feedback on the rps-auth draft (Cengiz Alaettinoglu)
IRR Secure Gateway (Ramesh Govindan)
Discussion of Work Group Charter (Curtis Villamizar)
Agenda Bashing (Cengiz Alaettinoglu)
Cengiz presented the meeting agenda. It was suggested that the
discussion of the work group charter be deferred to the end of the
meeting to allow extended discussion. This suggestion was accepted.
RPSL transition (David Kessens)
David highlighted the three phases outlined in the RPSL transition
draft. David then gave a brief summary of the status of the software
available. The software is:
BIRD (new development)
RaToolSet
RIPE registry software with RPSL extensions
RIPE-181 to RPSL converter
A few user trainings on RPSL have occurred. One was at NANOG, one at
RIPE, one is planned for APRICOT.
Phase II is now complete. We have software, it has been tested by
numerous parties, user trainings have occurred. Some providers are
into phase III already.
A meeting was held on Monday with developers and those known to be
running the registries or known to be actively working toward running
registries. Present were ANS, Cable and Wireless, the RA (Merit),
Telstra, Quest, ARIN, APNIC, RIPE, Connect. CANET-III and Bell Canada
were present by proxy (they described to David their status prior to
the meeting).
Brief statements of status were given at the RPS meeting by:
ARIN (?)
Looking at RIPE with RPSL extensions. Also looking at IRRd
but this seems less likely. The RIPE code based will be
modified to use an Oracle backend. A request was made that
they present performance findings at the next IETF. They hope
to be online in early 1999, maybe January.
RIPE (Joao Luis Silva Damas)
Using the RIPE perl code. They are rewriting the code from
scratch in C. The new code will support RPSL.
C&W (?)
Cable & Wireless is the former InternetMCI. They are running
older RIPE code with lots of private modification. They plan
to be ready by the dates agreed to at the meeting.
RA (Merit, Craig Labovitz)
Running RPSL in IRRd in a test database for 6 months now.
Migrating the RADB to ARIN is under discussion.
ANS (Curtis Villamizar)
RPSL capable configuration code had been tested by last IETF
but not deployed. Hoping to get real time mirroring deployed
in the same cutover but could transition now.
ISI (Cengiz Alaettinoglu)
RaToolSet 4.3.0 is out. BIRD is still under development.
It was agreed at Monday's meeting that by February 15 all of the
registries would have RPSL data available. At the next IETF we will
confirm readiness to shut off RIPE-181 after the IETF meeting in those
providers that are ready to shut it off.
RPSL Changes in New Draft (Cengiz Alaettinoglu)
All of the changes agreed to in Chicago and Los Angeles meetings are
in the current draft. Two additional changes were made. These are:
draft-ietf-rps-rpsl-extending-00 was merged in.
A leading "+" is allowed as an additional way to indicate line
continuation. This is in support of the ability to store objects
with blank lines, in particular PGP keys.
There were two requests for additional changes:
Add a formal grammar (requested at the last IETF meeting)
Add a list of changes from the previous version. A list had been
sent to the mailing list but did not appear in the draft.
Because there were changes, the RPSL draft must be recycled as a
Proposed Standard. After the two changes above are made a work group
last call will be repeated.
Feedback on the rps-auth draft (Cengiz Alaettinoglu)
BIRD now implements all of rps-auth except: reclaim, no-reclaim,
auth-override. The following comments on the draft have come from
this implementation:
There is no mnt-lower in the draft. RIPE had implemented this and
a need for this is seen. mnt-routes applies to routes only.
The syntax for mnt-routes is not explicitly stated.
It was suggested that the repository and delegated objects might
be candidates for moving to rps-dist since a central
implementation of rps-auth would not need them. The counter
argument is that rps-dist contains no extension to the RPSL
language, just contains the encapsulation needed for exchange
between repositories.
There were a number of more minor comments:
We don't really need a referal-in. This can be deduced by the
inverse query on referal-by.
Deletion of a maintainer is not allowed if the maintainer is
referred to in a referral-by. There was some discussion about
this feature. It was decided that it should remain as is.
Route authorization could fall within the hole attribute in the
route object. It was decided that hole is a comment and so this
should carry no implications.
If there is no aut-num, should the as-block be used for
authorization of a route addition? Currently the answer is "no".
It was decided not to change this.
The auth-override is effective within 60 days. This should be a
default, changeable on a per registry basis and reflected in the
repository object. If the maintainer has made any changes in the
last 60 days (or whatever the time period), then the auth-override
should not be placeable.
The default for reclaim/no-reclaim is currently implied to be
no-reclaim but should be explicitly stated and should be reclaim.
We don't have a no-delete to guard the lower in mnt-lower from
deleting the entry. Consensus was we don't need one.
The rules for as-set and route-set addition are defined in RPSL
and should be repeated in rps-auth.
IRR Secure Gateway (Ramesh Govindan)
There is a need to transition from the current set of routing
registries to a set of registries capable of rps-auth and rps-dist.
One method of transition might be to create a gateway between the
non-rps-auth/dist registries and the rps-auth/dist capable
registries.
There was some discussion but no conclusion as to whether there was
better ways to handle the transition to rps-auth. The gateway
provides an easy way to deal with the issue of non-rps-dist and
rps-dist capable registries.
Discussion of Work Group Charter (Curtis Villamizar)
The body in the new charter did not significantly change, particularly
the focus of the work remained the same. The milestones are brought up
to date to reflect the reality. New charter has three changes to the
old one, new wording of the first paragraph, reducing scope for tool
development, and changing the phrase "resource reservation" to "inter
domain quality of service". It was decided to keep the first two
changes, and drop the last paragraph altogether. It was noted by the
group and the AD that this did not stop the working group from
supporting QoS if and when it become important for inter domain
routing coordination.
The new RPSL internet draft will be recycled as proposed standard. No
more changes are expected (allowed) unless operationally absolutely
necessary. After 6 months, the WG will ask for draft status.
Do we need one of the following documents: rps framework and rps-dist
transition plan? Again, no decision was reached.
During the discussions, audience provided valuable feedback, it was
agreed that the RPS WG work was important for the operations of the
Internet. When asked, a surprisingly high number in the audience have
already read the documents and committed to providing feedback on
them.
It was discussed once the document specification work is over if there
was a need for a working group for deployment discussions. The answer
(thru clear show of hands) was yes. It was later asked if it should be
the RPS WG with a revised (deployment focused) charter or a new
WG. The consensus was to keep the RPS WG, but to finish with the
documents as soon as possible and focus on deployment. It was
suggested that a new WG would be disruptive. With the exception of
rps-dist, the WG committed to finishing document work by Minneapolis
IETF.
Curtis Villamizar will revise the charter as discussed and send it to
the WG and issue a WG last call that will end on January 7th. Cengiz
Alaettinoglu will add a section to RPSL highlighting the changes from
RFC 2280 and another providing formal grammar rules, and ask for a WG
last call to move it to proposed standard in December 1998. Cengiz
Alaettinoglu and Curtis Villamizar will incorporate rps-auth feedback
that was discussed and ask for a WG last call for a proposed standard
in January 1999.
