Next: Changes in the IEEE
Up: Results
Previous: Results
In [6], van Langevelde, Romijn and Goga relate experiences with the
tool Spin on Promela specifications of parts of net update for maintaining the
net tree and the bus trees, respectively. The most important result is the
discovery of non-termination of net update and a first attempt at the net
panic procedure. The error was detected with Spin simulation in the original
net update procedure, but not in subsequent fixes although these contained
the same erroneous behaviour.
In [8], Mooij and Wesselink use formal derivation methods
[1] to construct an abstract version of the net tree part of net
update. This is based on the correctness properties that must be obtained.
Currently, the proofs are being checked in PVS, while a front end tool for
such proofs to PVS is being developed.
In [7], Mooij, Goga and Wesselink construct an alternative
protocol for the net tree part of net update. The alternative is a mixture of
1394.1 and IEEE 802.1 (as described in [9,10]) functionality. The
correctness is proved with techniques from [1].
In [2], Goga and Romijn investigate the feasibility of
adjusting Promela specifications in order to restrict random simulations to
interesting behaviour. A theoretic approach is given, and it is shown that
given some sufficient conditions, the adjusted Promela code does not exhibit
new behaviour nor new deadlocks. Using this approach, the non-terminating
behaviour can finally be shown with Spin simulation in all erroneous versions
of net update.
A paper with manual proofs for the correctness of the net update and bus
enumeration protocol at the implementation level is in preparation.
Partial correctness has already been shown, termination still needs to be
proved.
Next: Changes in the IEEE
Up: Results
Previous: Results
Wieger Wesselink
2004-05-24