Τη Δευτέρα, 19 Μαρτίου 2012 3:50:35 μ.μ. UTC+1, ο χρήστης erik quanstrom έγραψε:
> > i'm not sure i understand the concept of reincarnation. on the one > hand, hardware by its nature can lock your machine up solid and > there's nothing the os can do about it. so how do you test driver > reincarnation? where could it crash? if you knew the answer to that, > you wouldn't need reincatnation. on the other hand, simple bugs > like dereferncing null i think would be better handled by fixing the > underlying problem. > > i wouldn't assume that nearly everyone else has taken a look at the > reincarnation idea. > > - erik Here's my rough explanation: AFAIU in MINIX 3 there is a server running (called rs) that "monitors" the rest servers (for example by sending small packets similar to keep-alives), and if one of them dies, it restarts it hoping that the problem that killed the server will not occur again. So rs might help in cases where a 3rd party driver crashes a server because of a bug, but it won't help if there's a critical deterministic error; as you correctly point out, the error needs to be fixed instead and there's nothing that we can do if we are talking about a hardware error. So I insist: I see no novelty in MINIX 3. It's just a clone of UNIX with the drivers moved in user space (something that can already be done in Plan 9) and the client-server model (which is already pushed much better in Plan 9), plus a reincarnation server (that can also be implemented in Plan 9). No 9P, no namespaces, no cross compilation by default, no clear text manipulation of device files, etc.... The fact that they have money and the EU picked them is irrelevant. The novelty of an OS lies in its technical aspects, not in the community or fundings...
