Minutes GSMP Working Group, Adelaide 3/30/00
--------------------------------------------
The meeting was co-chaired by Avri Doria and Kenneth
Sundell and was divided into two sections; the review of
current documentation and the discussion about future work
items for the working group.
The minutes were taken by Kenneth.
Changes to the three standards track documents currently
being worked on were discussed. While there were several
editorial and minor technical comments on the drafts, there
were no major technical objections. The plan is to update
the documents, fixing known problems by the end of April
and then to have a WG last call during May.
1. Review of draft-ietf-gsmp-04
Avri presented the changes since revision 02. Revision 03
was issued before the editing session in New Orleans (Feb
3), while revision -04 was released after that session.
Since the differences between 02 and 03 has not been
presented at an IETF meeting, though they had been
presented on the mailing list, the co-chairs decided to
present all changes from 02 to 04 in this session. There
will be at least a revision 05 before the document will
proceed to working group last call.
The encapsulation section was, as previously agreed, moved
to a separate document, draft-ietf-gsmp-encaps-00.txt.
All failure response messages were gathered and moved to
appendix A.
The 04 introduced the notion of TLV Labels, i.e. labels
that are longer than 28 bits. For clarity, the notion of
extended label has changed to stacked label.
The 03 introduced the reservation message which allows for
resources to be reserved for a connection before the actual
label mapping is known. Revision -03 also included support
for redundant adjacencies.
A proposal for adding output/input service selectors in the
Add Branch message had been discussed on the list and also
proposed at the New Orleans editing session. The capability
was added to 04.
Other additions to 04 included: a simple mechanism for
turning of the flow control, a way to do multicast queries
in order to support those switches that have dedicated
ranges of labels for multipoint connections, support for
discontinuous label ranges, the addition of event sequence
numbers and event flags to the port configuration message.
Additionally, the service model was updated to include
support for Circuit Emulation Services (-04).
2. Review of draft-ietf-gsmp-mib-01 (15)
Joachim Buerkle presented the updated GSMP MIB that is co-
authored by Hans Sjostrand and Balaji Srinivasan.
The GSMP MIB was totally rewritten. The GSMP MIB is focused
on the protocol and not covering partition issues. Joachim
pointed at work done in MSF where they are producing a
protocol independent partition MIB. He went through the
different GSMP MIB objects (slides will be available in the
proceedings).
There was a question about whether the structure of
multiple controller to a single switch partition should be
represented in the MIB. The answer was yes.
The document will be updated during April. Updates in the
next version will be mainly changes in wordings, some new
tables and maybe addition of statistics into the MIB.
3. Review of draft-ietf-gsmp-applicability-00.txt
Avri presented the applicability draft. There is a minor
set of documents that are needed before we can proceed to
IESG group last call. Apart from the GSMP specification,
GSMP MIB and GSMP Encapsulation documents, an informational
RFC for GSMP applicability is needed. The document
introduces the reader to concepts and notions that are used
within GSMP e.g. switch partitioning, controller/switch
interaction and service support.
A question about the MSF's role in relation to GSMP was
raised, i.e. where they defining their own version of GSMP.
Avri answered that IETF is defining the protocol and MSF
will be referencing the RFC in its implementation
agreement. Joachim clarified that MSF draft applicability
documents are not publicly available until they have been
approved for release by the membership.
The minimal set of messages needed to support MPLS was
discussed. The applicability document list the messages
required for minimal MPLS support.
4. Security issues in draft-ietf-gsmp-encaps-00
The main content of the encapsulation draft was moved from
draft-ietf-gsmp-02.txt. A security section has been added
to the document. A change that needs to be made to that
draft involves requiring the support of IPsec when the IP
encapsulation is used.
5. Working Group Last Call
Text revision of the GSMP spec will be due in approximately
3 weeks. The GSMP MIB co-authors will try to make a new
revision within the next 3-4 weeks. It is very important
that people read the documents and send appropriate
comments on the list. Non technical editorial changes
should be sent to the authors directly.
A two weeks WG Last Call will be issued in the beginning of
May.
6. Implementation Experience
Joachim Buerkle of Marconi communications (wireline access
system) presented on their work with GSMP. They have
produced I-Ds on resource reservation, move output branch
and GSMP MIB. In their implementation they are using the
ATM encapsulation and implementing the ATM specific
features.
There were questions about interoperability. A lot of
people say that they are implementing GSMP, but only a few
have done this in public. Service providers are soliciting
products.
7. Discussion of charter update
The group then spoke about future activities. Unless the
charter is redefined, the working group will go dormant
after the last call until implementations and
implementation issues crop up. A discussion was held to
determine whether there was real interest in adding
additional features to GSMP beyond those currently planned.
Several different possibilities where discussed.
7.1 Requirements for GSMP support of Optical switching
A presentation was made on using GSMP for controlling Edge
nodes, and optical interconnects within an IP over optics
environment. This presentation was essentially a
repetition of one given earlier in the week at the IPO BOF,
though it did focus more on the changes required in GSMP
then on the possibilities for optical network control.
Basically, a controller/router could control an OXC over
GSMP, either through use of an out of bound network or via
a dedicated "control" wavelength.
New label TLVs were introduced to handle label types such
as fiber bundle, arbitrary number fibers in that bundle, a
single fiber, arbitrary lambdas in that fiber, a single
lambda etc.
Several questions were raised; whether we need a raw
optical encapsulation for GSMP messages or if we should
require use of the IP encapsulation, whether we need one
single TLV or several for accommodating the different types
of optical labels and whether there was an existing service
model for optical switches which GSMP could use or would
the group need to define one.
7.2 Future Requirements for GSMP
Several other proposals for new work were discussed:
- Add support for SDH port types.
- Investigate a rework of the flow control mechanisms in
GSMP.
- Add support for switching IP packets (L3 switching).
Part of the support for this already exists in the
protocol, but many of the elements which would be
required, are still lacking. This effort would probably
require a requirements effort first.
- Continue tracking work on QoS service models.
7.3 Conclusion
There was definite interest in continuing the work of the
WG. In fact instead of having a passive (i.e. does anyone
object to continuing work on these and similar items), we
had an active show of hands on whether folks thought these
issues were worth pursuing. Of those voting, and
approximately 2/3 voted, there was a clear consensus for
continuing the group. The intention is the chairs discuss a
specific charter amendment on the list before making the
formal request to the IESG.
In terms of vendors and service providers interested in
seeing the work continuing, there was positive interest
from several companies.