
It's been a couple of months since I read the systems manual for 787 (given current delays, learning to drive the thing has lost the urgency!), but while I think some of the concepts didn't look brilliant on paper I don't think anyone would get far trying to get into the avionics from a cabin facing system.
Much of the feature set of the aircraft does smack of the design team throwing every idea they could come up with at the thing, in some cases ending up with features apparently added just because they could be! Plus they've moved from the traditional avionics concept to something more familiar in the commercial world i.e. file servers, processing servers, and switched networks.
As far as the computers goes, I seem to remember that on top of the avionics RTOS, there are also instances of Linux in there, and even XP (!!) as part of the Electronic Flight Bag modules (not a concern though, it's only used for the HMI, and the aircraft data exchange is handled via another OS (Linux?) on an independent CPU i.e. XP talks to Linux, Linux talks to aircraft. So Microsoft gets nowhere near the actual avionics).
I think where most of the reported concern comes from the way the back end hardware is managed.
The actual aircraft avionics networks run CDN (ARINC 664), over fibre. This is basically the same thing as used by Airbus on the A380 (AFDX), though in that case traditional cables are used instead of fibre. While this is derived from Ethernet it isn't the kind of thing you'd easily patch into or upset with a laptop. The cabin applications network is for the most part independent from this (dedicated cables, different protocols), with interconnection only where required to share common hardware platforms or for data exchange.
The networks do come together in the core network and the Common Computing Resource. The former is the shared part of the aircraft networks, and incorporates all the network switching required to interlink various subsystem networks. The latter is the computing resource for the avionics and other systems, and can basically be considered as a form of blade server. The concept is that all the avionics processing runs on a common hardware platform, with the various systems being allocated to modules as required. This makes spares and support easier, and allows simple fail over during operation. In the worst case complete failure is handled by dual redundancy of the complete CCR, networks etc. etc., and beyond this low capability fail-safe modes in individual subsystems e.g. in the case of complete failure of the networks.
Anyway, I know I didn't think much of the idea of all this concentrated and connected hardware when I first saw it, especially given there are wireless networks and cabin accessible networks involved.
But it should be remembered that the way the system is built, with isolation and firewalls, should be adequate - after all, the different networks are only linked in that they share a switching platform at one point, and run different protocols and it's quite possible to ensure that only specific, formatted packets can be transferred between specific points, and even then only in one direction. I would assume that there isn't actually any flow from one network directly to another, but that certain systems have a link to the avionics net and also a dedicated port for any outputs to other networks e.g. cabin systems or maintenance.
Considering that the hardware is all custom, the operating system is custom, the critical networks use a robust avionics network protocol and that the concept that someone might cause a problem (deliberately or otherwise) is a fairly basic one - let's face it, in theory a faulty IFE system could cause just as many problems on the network! - I don't think there was ever any real likelihood of a problem as it's been considered at a low level from day 1, for example in the design of the network protocol. When you have complete control you can build the systems to do specific jobs, and no more, which makes them slightly more robust than a general purpose operating system with a fully featured network stack.
Not that I'd really want to go on a 787 for the first 12 months after entry into service, or live under a flight path: there are just too many things that 'might' go wrong given how many of the concepts are new. I'm yet to be convinced that composite fuselage tube is completely safe on an aircraft this size...