SPJ http://research.microsoft.com/en-us/people/simonpj/default.aspx and probably many others are actually employed at Microsoft research centers. It looks like Microsoft just hasn't been able to find a suitable spot to push Haskell. Haskell influenced F# because they needed a functional language that targeted CLR, and included OO and mutable data. IMO Haskell is even better than their languages

Maybe so but singularity actually provides the whole os apis via clr interfaces compared to mainstream windows os where the underlying apis are all in C, C++ and COM. The common intermediate language is not tied to any specific programming language such as C# or VB, it's more generic than that, and has it's advantages. Safety is something they wish to achieve but afaik their main goal is to write an OS in managed code. Haskell does provide a safe runtime but afaik unlike the clr it's tied to the haskell language. I think there has also been some attempts to write an OS in haskell too though, but that's another story...

Haskell's great and all but it does have a few warts when it comes to how much real trust one should put into the type system. Some compromises still exist like unsafePerformIO that you can't detect simply by looking at the types of functions. In order to live up to the hype and the marketing around Haskell, really things like unsafePerformIO should not be allowed at all. The type of unsafePerformIO $ fireTheMissles >> return 3 ::Int is just Int after all. Does Singularity also have such back doors? Dave On Sat, Jul 31, 2010 at 6:53 AM, Alberto G. Corona wrote:

I guess that the house OShttp://www.google.com/search?hl=en&safe=off&q=+house+OS+haskell&aq=f&aqi=g-sx7&aql=&oq=&gs_rfai=has no one of these problems that singularity tries to solve in the first place.

The problem of general OSs is: we have unsafe code, so what we do to deal with it?. The usual option is the isolation trough virtual addresses so that every pointer address is virtual. This imposes cost in task switching and pointer handling. The singularity alternative seems to be to check the managed code for pointer violations at installation time.

In singularity they pretend to extend the reach of types, defined in .NET at the assembly level for inter program and inter language safety, to the OS level for runtime safety. This goal is interesting, because a well defined type system, without unsafe operations permitted, managed at the OS level could permit pure code to run wildly in real memory very fast, for example. With effects defined in the type system the advantages may be greater.

2010/7/31 Tim Matthews

...

SPJ http://research.microsoft.com/en-us/people/simonpj/default.aspx and probably many others are actually employed at Microsoft research centers. It looks like Microsoft just hasn't been able to find a suitable spot to push Haskell. Haskell influenced F# because they needed a functional language that targeted CLR, and included OO and mutable data.

IMO Haskell is even better than their languages

...

Maybe so but singularity actually provides the whole os apis via clr interfaces compared to mainstream windows os where the underlying apis are all in C, C++ and COM. The common intermediate language is not tied to any specific programming language such as C# or VB, it's more generic than that, and has it's advantages. Safety is something they wish to achieve but afaik their main goal is to write an OS in managed code.

Haskell does provide a safe runtime but afaik unlike the clr it's tied to the haskell language. I think there has also been some attempts to write an OS in haskell too though, but that's another story...

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

David Leimbach wrote:

Haskell's great and all but it does have a few warts when it comes to how much real trust one should put into the type system.

Some compromises still exist like unsafePerformIO that you can't detect simply by looking at the types of functions.

In order to live up to the hype and the marketing around Haskell, really things like unsafePerformIO should not be allowed at all.

As I mentioned in the thread about escaping monads, you actually have a proof obligation in order to use unsafePerformIO. The only problem is that those obligations are not captured in the source language itself, so you must trust the code you link against, separately from any trust induced by type checking. There are very real reasons for wanting a function that can take an IO A into A, which is why unsafePerformIO was added in the FFI addendum. The only way to correct this situation is to (a) add a proof theory to the Haskell language, a la dependent types; or, (b) to break apart the IO sin bin so that we can track the more innocuous parts independently from launching missiles. Of course, the second approach also requires proof that information from the, e.g., RTS monad does not leak into the return value of runRTS. To do this in general without loosing the power we want from RTS, we'll need to add a proof theory to the language in order to demonstrate that two functions are extensionally equal. So really, the first option is the only one; in which case you might as well switch to Agda or the like. -- Live well, ~wren

And note that we wouldn't need unsafePerformIO for the FFI if all programs were made in Haskell ;).

Perhaps that's true, though entirely unrealistic, in the application world. In the OS world you need access to machine registers and special instructions (CR3 anyone? CP15?) which isn't built into any language save assembly - for these FFI will always come in handy. Also, Haskell continues to have an unfortunate lack of primitives suitable for casting types (ex: zero copy form a bytestring like entity to Word32s). In this realm FFI can outperform cleaner looking code that must rely on individual byte reads. Cheers, Thomas

On Sat, Jul 31, 2010 at 8:27 PM, wren ng thornton wrote:

Thomas DuBuisson wrote:

...

...

And note that we wouldn't need unsafePerformIO for the FFI if all programs were made in Haskell ;).

Perhaps that's true, though entirely unrealistic, in the application world.  In the OS world you need access to machine registers and special instructions (CR3 anyone? CP15?) which isn't built into any language save assembly - for these FFI will always come in handy.

Also, Haskell continues to have an unfortunate lack of primitives suitable for casting types (ex: zero copy form a bytestring like entity to Word32s).  In this realm FFI can outperform cleaner looking code that must rely on individual byte reads.

The FFI doesn't always require unsafePerformIO,

True. I mis-read the previous e-mail as "we wouldn't need unsafePerformIO OR (vs for) the FFI " so please ignore that response to a non-existent statement!