Diagnosing a Stuck Growing Point

Diagnostic services localize the source of the failure and characterize the form of the breakdown. When a well-formed GPL expession fails, we may assume the mode of failure is a "stuck" growing point. As we mentioned in the last section, the only reason a growing point becomes stuck is because none of the neighboring sites matches the criteria for propagation. These criteria may not be met for a variety of reasons.

Sparseness:

There are an inadequate number of processors to choose from.

Unsuitable pheremone values:

After the tropism expression filters through the neighboring pheremone values, there are no suitable successors.

Noisy channel:

Processors are all trying to communicate simultaneously, and too many messages are colliding. This can produce spurious results because incomplete information will alter the growing point's perception of the local landscape, causing it to propagate suboptimally, or not at all.

Poisoned processors:

Neighbors that have become poisoned will specifically ignore REQ's for certain pheremone values. This altering of the geometry with respect to the growing point may be intended by the designer, so the poisoned processors should be respected.

In all theses situations, the protocol to determine the cause of failure may be initiated by the stuck growing point, and implemented in ECOLI.