Slashdot Mirror
Operating Systems of the Future
An anonymous reader writes: "'Imagine computers in a group providing disk storage for their users, transparently swapping files and optimizing their collective performance, all with no central administration.' Computerworld is predicting that over the next 10 years, operating systems will become highly distributed and 'self-healing,' and they'll collaborate with applications, making application programmers' jobs easier."
IMHO, future operating systems will tend to something like the ErOS operating system . This OS is based on multiple tiny extremely reliable components, within a strong capability model to provide a high level of security.
It's definitely a good approach, although ErOS is still quite experimental yet.
{{.sig}}
I predict that there will never be a revolutionary new operating system until we break free of the chains imposed by Posix compliance. Until then, we're stuck with files that have to be streams of bytes, ugo-style permissions, non-wandering processes, incompatable RPC calls, &c.
And the real pain is there have been OS'es that have had simple & elegant solutions to problems that are hard under unix (Aegis, Multics, VMS, TOPS, ...) that were pushed aside by the steamroller that is Unix.
But to be fair, many of the forgotten O/S's are now forgotten because they weren't as general purpose as Unix. Unix is the great compromise. But it's hard to strive for the best when you've already accepted compromise.
"Imagine computers in a group providing disk storage for their users, transparently swapping files and optimizing their collective performance, all with no central administration."
Whoever thought up this pipe dream apparently doesn't understand the Zeroth Law of Network Security: If you want information to be secure, DON'T PUT IT ON THE FUCKING NETWORK!
Seriously! As if most business OSes don't default to the least-secure settings already! Why would you want to run important apps on a system where the default is to share anything and everything with any computer in listening distance?
...there won't be much drastic change from now till the next 18 years. For evidence of this, look at the Apple Lisa. The Lisa had windows, icons, a menubar, a WYSIWYG interface, and a mouse. Today's computers are little more than a glorified Lisa interface, whether they are running Mac OS X or Windows XP (I know because I run both.) Like the Lisa, todays computers still crash and still corrupt themselves. I doubt that this could be easilly changed in the next five, ten, or even fifteen years.
I'll believe the distributed file-storage myth when I see it. To me, it sounds as if it would hog bandwidth, just like gnutella does. I don't see any change coming in the way I store files on my computer. It's fast, effecient, and hasn't needed a change.
SysAdmins need not quit their day-jobs. As long as Microsoft is providing this technology, you can be sure that it will run into snags and security vulnerabilities. Increased complexity = increased vulnerability.
...and that's all I've got to say about that
My strong belief is that the best "predictions" occur when you find something in use today - only too expensive for the home user - and "predict" it will be ubiquitous within a few years. So here are my completely predictable predictions.
Notice how all of my predictions sort-of exist already. This is what makes predictions so easy.
Sure, you've got to start with reliable components, but you have to combine them in just the right way, too.
First off, we should learn a lesson from biology. The bee, for example, has about a million interconnected neurons. Yet the bee's highly sophisticated behavior is extremely robust and efficient. How does nature do it? The answer has to do with parallelism and expectations.
1. Parallel processing insures that signals are not delayed, i.e., their relative arrival times are guaranteed to be consistent.
2. Expectations are assumptions that neurons make about the relative order of signal arrival times.
We can emulate the robustness of nature by first realizing that computing is really a genus of a species known as signal processing. We can obtain very high reliability by emulating the parallelism of nature and enforcing a program's expectations about the temporal order of messages: no signal/message should arrive before its time. The use of stringent timing constraints will ensure that interactions between multiple tiny modules remains consistently robust. Enforcement should be fully automated and an integral part of the OS.
Of course, this is only part of it. The other constraints (e.g., the use of plug-compatible links, strong typing, etc...) are known already. No message should be sent between objects unless first establishing that plugs are connected to compatible sockets, i.e., that they must be of the same type.
The most problematic aspect of computing, IMO, is that it is currently based on the algorithm. Problem is that algorithms wreak havoc in process timing and the end result is unreliability. The algorithm should not be the basis of computing. To ensure reliability, computing should be based on signal processing. Algorithms should only be part of application design, not process design. Just one man's opinion.