When the software meets the hardware
On 21 Jan 2019 11:38:48 +0000 (GMT)
Theo wrote:
In uk.transport.london Recliner wrote:
Yes, that is surprising. Presumably the same data feeder system is used,
but the computerised visual and audio subsystems are different, maybe even
from different sub-contractors.
I assume it's like a modern car, which is a distributed system containing
dozens of ECUs (ie computers) flying in loose formation, joined by a
network. In the case of an 8 or 12 coach train there are probably hundreds
of nodes.
Building distributed systems is hard, especially when heterogenous, and when
Not really. So long as there is a published API/interface to each subsystem
then the seperate nodes should just be black boxes with internals that the
system intergration team shouldn't have to worry about. The problems arise
when the published interfaces and/or behaviours don't match the actual ones.
involving physical inputs which are difficult to simulate in a test
environment (eg the kinematics of the train, doors, toilets, etc).
Toilets don't need to be software controlled in the first place. Only teams
trying to justify their jobs would make them so.
|