Transition and Coexistence Including Testing BOF (TACIT)

Reported by Atul Bansal/Digital and Geoff Huston/AARNet


Introduction

Geoff Huston presented issues related to the deployment of a new
technology within an existing environment.  The key points of his
presentation were:


   o The end-user requirements drive the transition to a new technology.

   o The deployment of the new technology is purely driven by the
     transition cost and the benefits of the new technology.

   o Testing of a new technology is mandatory before it can be deployed.

   o The objective of the testing is to gain confidence with the new
     technology.

   o Various testing methods such as test beds, bake-offs, simulators,
     etc., could be used to ensure functional stability,
     interoperability, performance, manageability, etc.

   o The cutover deployment model will not work for Internet.  The new
     and old technology will coexist for a long period (as long as the
     current machines exist in the Internet).

   o The deployment will be incremental.

   o The deployment/transition plan must provide autonomy; independent
     transition of the network service, site, host, and applications.

   o It is not practical to specify any ordering on the transition,
     however, dependencies must be listed for each transition mechanism.


Given that it is now known that SIPP-16 forms the basis of the IPng
recommendation, the proposed TACIT Working Group charter was reviewed.


Discussion

The TACIT BOF noted the distinction between the TACIT activity and the
IPng transition activity in that the specific transition issues
regarding protocol transition will be defined and managed through the
IPng transition activity, while TACIT will provide the testing
(conformance, assurance, interoperability and performance) and
deployment expertise generic to the transition.

The TESTING BOF also has areas of potential overalp of activity, and
again the envisaged scenario is that there are generic guidelines
relating to testing which can be applied to a process of testing any
specific architecture.

The TACIT BOF envisaged that the protocol correctness testing for IPng
would be undertaken by the designers and implementors of the protocol.
They would also be responsible for testing scaling of the protocol.  If
there were to be other related transitional activities (such as IPX to
IPng, or CLNP to IPng) then TACIT would be working within the area of
generic guidelines for such transitions, and these would be applied to
specific transition plans undertaken within related, but distinct,
activity domains.

It was noted that the planning for the IPng transition and the TACIT
work would be commencing at the same time, and there are noted issues
within TACIT providing input into the IPng transition plan, and there is
a requirement to attend to coordination issues between the two groups.
The detailed IPng transition plan may be reviewed by TACIT in general
terms, but the responsibility for the actual detailed transitional
structure is the IPng transition activity rather than TACIT.

One of the deliverables is to be a ``lessons learned from other
transitions'' document, where the goal of TACIT is to learn and document
the assembled information about deployed protocol transition.


Presentations

Initial IPng transitional concepts were presented by Bob Gilligan.  The
BOF noted the problems inherent with the use of tunnels within
production service platforms, and also discussed whether multicast
techniques were viable in such an environment.

Dave Piscitello presented some generic issues concerning transition
noting that transitions which were gradual, feature driven and used the
minimum level of interdependencies were generally more successful.  The
critical role of the DNS in providing information to the complete
transitional protocol environment was stressed.  The role of translators
was noted as a likely component of a transitional environment.