Talking about processing power (I love this thread, lots of ideas boiling up), the decks are said to use some limited form of parallel processing to gain enough power to bypass the security measures. Then I think it would be reasonable to say that the hardware took a general powerup with the new new matrix for the sake of handling these protocol, right?
Which means that even commlink have to have increased in power themselves. Not to the same extent as they don't perform the same tasks than decks and use legitimate keys and so on, but if they are keys and so on (because keys are an old concept, security protocols evolved in nobody knows what) that need the power of multiple links to be broken, I guess that mean their use would no be so transparent when it comes to strong protocols (like the equivalent of AES 256 or RSA 2048 and things like that for our current computers).
If you think about it, there could be "levels" of bonuses when they come to be possibly related to distributed computing.
Interfacing your stuff with your commlink (by any mean, which is more convenient for you) would give the least, at best a +1 bonus on a test limit, and some special goodies like ignoring a normal glitch involving the use of the implant for the most powerful of links, interfacing with a deck would give a +2 to test limit, plus ability to ignore a normal glitch and maybe a few things not measured in dices and ratings, and being connected with a mainframe or the matrix would go so far as give a +1 to +3 dice pool bonus in addition of the rest, for example.
Granted it would make sense. But for such applications I think I would require special additional programs like "pilots" to run on your link to interface the thing properly with matrix. No matter how efficient are the protocols, the facts that implants are very specific and these tasks somewhat not made to be handled in an internal fashion would mean to me that you need something specific to handle that.
To balance it, I would take noise into account when connection woud be needed. And maybe the fact that distributed computing or not, at times calculations take a so specialized form that they can't be divided in smaller parts than a given threshold. I thinking for example of some special matrices types that should only be handled by using multiple dynamic operations at the same time. I don't know if that actually exist but I know there are some programmation objects in some languages that can only be managed in specific conditions, in dynamic or static ways (and I am not talking about the object type, but about its internal processing). Doesn't seem odd to me to imagine that it would exist for some tasks (like encryption), and then you would not be able to take advantage of being in an electronic shop because those puny comlinks or below would not be able to handle these smallest parts from the start.
That might be something I would houserule given time.
Still agree with Railgun on the fact that sometimes, too much power is not needed.
Distributed computing doesn't need a dedicated server if the networking protocols include the ability for every device to be its own central server for each task. That reduces the algorithm from quadratic to linear. </nerdtalk>
Well, I don't really know that kind of things and it's a pity, but from what I understand you are talking about the algorithm giving the task execution time depending on the time of protocol? It is very counterintuitive, or maybe I did not understand the concept of "every device being its own central server for each task.", because for me, if there are multiple things deciding on their own what part they will take of the calculation for example, it will just end with redundancy and as such, with lost cycles. Also, could be data collision, since even if the calculation on part 1 is done, some other host could recalculate it again after. But that would not be. I guess my assumptions are plain stupid there because that could not be something some computer head would come with the idea of, but then I missed something on this concept.
I should take some more degrees in math and computer sciences. Learning by yourself hits limits at times.
@RHat: By IDA, do you mean that:
http://proceedings.spiedigitallibrary.org/proceeding.aspx?articleid=959791 ? Interesting. Thank you very much, wasn't aware of that.
Edit; Another idea came up which would fit with giving a role to deckers into fight; while my ideas allow the use of wires for some applications, deckers could still try to hack the deck (as usual), or instead, locally work the matrix in a way it would refuse distributed computing operations from some "IP" or "mac adress" (see, the SR5 equivalent of it). Or completely wreck the local grid in a limited way indeed, but perturbating the cooperation among the nodes. Indeed, something GOD would be very pissed off at. But sometimes it's better to be bold than to be dead.