Singularity?

[unixronin]

That's the codename for Microsoft's new research OS.  (Well, I say "new" ... they've been developing it for "more than five years".)  It's written in Microsoft's C#, which they claim prevents the possibility of buffer overruns.

"Singularity is not the next Windows," Rick Rashid, senior vice president of Microsoft Research, said in a statement.  "Think of it like a concept car.  It is a prototype operating system designed from the ground up to test-drive a new paradigm for how operating systems and applications interact with one another.  We are making it available to the community in the hope that it will enable researchers to try out new ideas quickly."

Could be interesting.  (Shame the architecture diagram in the article is too small to read, though.)  Some of the features they've designed in could potentially be very good ... but they also could be very bad.  (I might mention the registry as a well-known example of a good-sounding idea gone bad.)

Comments

[dafydd.livejournal.com]

Your link is an email address.

??

[unixronin.livejournal.com]

Oops. Not sure how that happened.

Fixed.

[ilcylic.livejournal.com]

"Singularity: Because it sucks hard enough to bend light"

[johnkzin.livejournal.com]

It sucks so hard, it bends even the best idea.


It's so bloaty that its mass bends space-time.


It sucks so bad, good ideas go in and are never seen again.

Drawing on that other type of Singularity

[metahacker.livejournal.com]

"Singularity: cause the end of the human race as you know it"
"Singularity: so advanced, you can't possibly comprehend it"
"Singularity: rapture for geeks (who haven't discovered a real OS yet)"

[cipherpunk.livejournal.com]

ObDisclosure: I am technically an employee of Sun Microsystems. My day job involves evangelization for Java and Solaris. My agreement with them is that I only say things I really believe--I am not a marketroid.

That said, forgive me for howling with laughter at the idea C# prevents the possibility of buffer overruns. Java, maybe, due to some features of its design. Once you start writing JNI to deal with low-level devices, though, all bets are off. C# has the exact same problem (through P/Invoke), but wait, it gets worse: to allow languages like C/C++ to run on the CLR, Microsoft explicitly allows code to deal with pointers by marking code blocks as "unsafe".

How much do you want to bet unsafe code is used in Singularity? How much do you want to bet low-level drivers are done with heavy use of P/Invoke calling out to unsafe languages?

[unixronin.livejournal.com]

I don't take sucker bets. :)






(footnote: there's something weird going on with LJ the past few days that's causing just-posted comments to not show up immediately, sometimes making me think the comment didn't post....)

[genuine-snark.livejournal.com]

I thought the secret sauce with Singularity was intra-module communication through secured channels.

re: the unsafe thing mentioned above, I believe once you have unsafe parts, you're no longer "managed." Since they say the whole thing is managed code, it may not be as hackish as you think.

Also, if you come into this with anti-MS bias, consider that this comes from the Research Labs. They do lots of neat stuff there, and run under near-complete autonomy from the rest of the company from what I understand. They're the think-tank portion of MS.

[unixronin.livejournal.com]

consider that this comes from the Research Labs. They do lots of neat stuff there, and run under near-complete autonomy from the rest of the company from what I understand. They're the think-tank portion of MS.

Yeah. As I said, it's interesting. I just don't buy some of the hyped statements.

There's also the aspect that "based in paradigms developed in the 60s" also means "based on paradigms that we now have 40 years experience at fine-tuning."

[cipherpunk.livejournal.com]

Point taken re: managed code--however, given I've seen "managed code" that did all of its heavy lifting by P/Invoke, I have my suspicions that this distinction is not one that is particularly well-honored in the real world.

[genuine-snark.livejournal.com]

Well, my basic thought is that if part of the success was making everything managed, it'd be a hell of a compromise to have unsafe code running around in there.

That's not to say that they didn't, but usually that kind of compromise comes with a downplay. Also--and I have nothing to back this up--I think the Research division would be more likely to actually "do it right." Production code's a different story.