CURRENT_MEETING_REPORT_
Reported by Sue Hares/MERIT
OSINSAP Minutes
Agenda
o Introductions
o Status of pending RFC:
o ``OSI NSAP Address Format for Use in the Internet''
o ANSI Registration for NSAPs
o Review of: ``Guidelines for OSI NSAP Allocation in the Internet''
Status of NSAP Structure RFC
Ross Callon reported that the RFC has been reviewed and approved by the
IESG. However, the IAB approval is held pending additional descriptions.
The IAB seems to desire the solution to all possible problems with the
ISO addressing format prior to approving the document as an RFC.
Few people had obtained the last copy of the document. Ross Callon read
the guts of the document. Richard Colella solicited comments. Juha
Heinanen suggested some corrections in the sentences regarding European
additions. Richard collected all the comments and will re-publish the
document by the 9th of January. All comments should be into Richard
with the last weeks of December.
ANSI Registration
People can now obtain organization IDs from ANSI for use in NSAP
addresses. ANSI assigns organization IDs for NSAPs that have the ISO
DCC format and the United States country code. ANSI currently is only
registering the numeric form of the organization ID. Registration of the
alphanumeric form is expected in the first quarter of 1991.
The fee for a numeric organization ID is $1000. Assignment of a name
will be made within 10 working days. Previously, ANSI had a queue of
800 requests for organization IDs. ANSI will ask all these people to
re-apply using the new procedures. ANSI expects the re-application to
happen in a manner that will allow them to maintain their 10-day
turn-around time.
1
A copy of the application form was available at the meeting. Anyone
wishing a copy of the form or other information regarding ANSI
registration can contact ANSI.
ANSI
Organization Name Identification Code Assignments
1430 Broadway
New York, NY 10018
voice: (212) 642-4976
fax: (212) 302-1286
Review of ``Guidelines for OSI NSAP Allocation in the Internet''
Ross Callon gave a general overview of the paper ``Guidelines for OSI
NSAP Allocation in the Internet''. People who had attended the ANSI
X3S3.3 Working Group noted that ANSI had elected to suggest a DSP format
for the ANSI DCC code that was identical to the GOSIP 2 format. (This
format is the one selected in ``OSI NSAP Address Format for use in the
Internet,'' RFCXXX.)
The ANSI format under the US DCC would be:
| AFI | IDI | <-- DSP --> |
-------------------------------------------------------------------
| 39 | 840 | ORG ID | DFI | Rsvd*1 | RD | Area | ID |sel|
-------------------------------------------------------------------
No. of bytes: 3 1 2 2 2 6 1
This DSP format is identical to the GOSIP 2 format.
*1 - GOSIP calls this field `Reserved'. However, `Reserved'
has a different meaning in ANSI than as used in GOSIP. In both
cases, this field needs to be set to a par- ticular value and
the users need to ignore the value for now.
The DSP Format Identifier (DFI) allows alternative DSP formats to be
defined by ANSI in the future (this is identical to the DFI field in
GOSIP 2).
2
After the basics had been covered, the NSAP Working Group spend a great
deal of time discussing issues of assigning NSAPs to three different
types of Routing Domains:
1. Zero homed - routing domains not attached to anyone.
2. Single homed - routing domains only attached to one regional
network.
3. Multi-homed - routing domains attached to several regional
networks.
The ``Guidelines for OSI NSAP Allocation in the Internet'' proposes a
carrier-based NSAP assignment plan. Many people attending the Working
Group wanted to see this contrasted with a geographical based NSAP
assignment plan. Ross and Richard lead a discussion of how each of
these types of routing plans work for the three types of Routing
Domains.
Due to the richness of the discussion, the note taker could not capture
the full discussion. I've attempted to capture some of the discussion
below. If I've missed somone's comment, please send the additional
information to the mailing group.
Discussion of NSAP Allocation
Richard described a zero-homed routing domain as:
1. No connections into regional networks.
2. Private point-to-point links using leased lines or dial-up used as
unadvertised back-door links.
3. Routing information is not sent to the rest of the internet
(essentially, an isolated Routing Domain).
Single Homed Routing Domains
1. May have multiple links into a regional network.
2. Only attache to one regional network or directly to one national
backbone.
Discussions on the actual status of regional networks broke into richer
descriptions of the types of routing domains:
3
The phone companies use a phone number based on local carrier. It seems
to be geographical due to the structure of the phone companies. Ross
Callon suggests that the geographical nature of the phone system is
simply due to the fact the phone company maps its logical topology onto
a physically geographic topology. It is the logical/carrier-based
topology that is really being used.
[A great deal of discussion centered on this point.]
Regional networks are not geographic in nature. Sue Hares noted the
case of the state of Idaho where half of the colleges are served by
Westnet and half by Northwestnet. The reason for the split was the high
cost of the inter-state phone lines.
It was noted that geographically-oriented routing may tend to create a
flat space of routing domains, rather than a hierarchy of routing
domains.
Vint Cerf noted that this discussion of geographical versus
carrier-based has been a long-standing discussion dating back some 25
years. A mid-ground in the discussion might be using the classic idea
of default:
1. If you don't know where to send it, push it up the hierarchy.
2. Hierarchical knowledge puts the burden on the national networks who
have more resources.
Vint Cerf also asked that any allocation plan try to look at the sources
and sinks of traffic.
Juha Heinanen noted that we were talking about three alternatives:
1. Flat data space for NSAP - such as the Internet has.
2. Subscription (or carrier) based addressing.
3. Area Code space.
Ross Callon noted that use of the geographical naming has extreme
problems when a national corporation connects to three different
carriers. The national corporation may want to send traffic to the
nearest exit to their private network which spans the United States.
Guy Almes cautioned that we must not confuse explicit route with a
4
particular Address format.
Phil Almquist brought up the idea of a default carrier so the national
corporation would default to a particular carrier.
Vint Cerf indicated it might be fruitful to look at how ISDN selects a
terminating host. The use of IP in the ISDN world brings up issues that
may have some bearing on the Internet.
As time was running out, Richard tried to gather specific changes to the
NSAP guidelines document. The following are my collection of changes:
o Add information about the zero-homed routing domain.
o Add more about multiple links into a single homed routing domain.
o Possibly put in an appendix a list of unanswered issues.
o Put in examples using real life network topologies.
o Indicate how this type of NSAP allocation will support future
changes to the Internet. Guy Almes indicated that the structure of
regional network may change.
o Ross Callon's example of how a NSAP prefixes work in each of the
three cases for MEGA Big Incorporated.
A separate paper on geographical versus carrier-based OSI NSAP
allocation was suggested. The IAB needs some description of these
issues if it is to discuss them. Such a paper would focus on the pros
and cons of each type of NSAP assignment. It would need to examine past
work on the subject, current topology and future needs. There were no
volunteers to author this paper.
Attendees
Steve Alexander stevea@i88.isc.com
Guy Almes almes@rice.edu
Philip Almquist almquist@jessica.stanford.edu
William Barns barns@gateway.mitre.org
Ross Callon callon@bigfut.enet.dec.com
Lida Carrier lida@apple.com
Vinton Cerf vcerf@NRI.Reston.VA.US
Richard Colella colella@osi3.ncsl.nist.gov
Curtis Cox zk0001@nhis.navy.mil
Steve Deering deering@xerox.com
Dino Farinacci dino@esd.3com.com
Debbie Futcher dfutche@relay.nswc.navy.mil
5
Martin Gross gross@polaris.dca.mil
Robert Hagens hagens@cs.wisc.edu
Tony Hain alh@eagle.es.net
Susan Hares skh@merit.edu
Juha Heinanen jh@funet.fi
E. Paul Love Jr. loveep@sdsc.edu
Andrew Malis malis@bbn.com
David Marlow dmarlow@relay.nswc.navy.mil
Tony Mason mason+@transarc.com
Cyndi Mills cmills@bbn.com
Daniel Molinelli moline@trw.com
James Mostek mostek@cray.com
Mark Needleman mhn@stubbs.ucop.edu
Fred Ostapik fred@nisc.sri.com
Theresa Senn tcs@cray.com
Keith Sklower sklower@okeeffe.berkeley.edu
Linda Winkler b32357@anlvm.ctd.anl.gov
Dan Wintringham danw@osc.edu
Cathy Wittbrodt cjw@nersc.gov
Richard Woundy rwoundy@ibm.com
6