Lambda the Ultimate

I came across this paper a couple of months ago,a nd remember discussing it. I thought it was here, but I can't find it. I'll try to dig up what I wrote then.

I came across this paper a couple of months ago,a nd remember discussing it. I thought it was here, but I can't find it. I'll try to dig up what I wrote then.

Doing a quick search, I found that this thread. I updated my Lambda threads sorted by topic and sorted by date which helps me when I want to look up some previous discussions/links.

As for the paper, I agree that Operating Systems research has gone into remission. Even new OS's being marketed these days (BeOS and OSX come to mind) are just built on top of the same old Unix shells. The mach kernels are going no where fast (Next is all but dead, and the Hurd is being crushed by Linux).

That said, I think the opposite is true for programming languages. There have been numerous programming languages come out in the last ten years, and I don't see any let up in the further expansion. New languages crop up every year and the number of languages that the average programmer uses is now above 2-to-3.

I agree with some of what Pike is saying, but I also can't help but wonder if some of what he says is fueled by Plan 9's lack of acceptance in the marketplace.

Systems research (as Pike pointed out) requires a long-term commitment, as well as some considerable funding. It is not surprising that people with specific interests (scheduling, real-time systems, etc.) will take something with freely available source code like Linux, BSD or Mach and alter it rather than building from scratch.

But there is some good work being done. For example, there are Sony CSL's Apertos OS, and the University of Washington's Spin OS. Also, there is work on using partial instantiation to speed up system calls; this work is funded by HP, and (I think) it was being done at the University of Oregon by Carlton Pu.

The real problem here is the computer industry. Few companies write their own OSes, and it isn't in their interests to throw out an existing codebase with years of work invested in order to field a new, mostly untested OS. It is not in the interest of developers to retrain themselves on new systems with possibly unfamiliar APIs when they could be writing for known, established plaforms.

(An example of this can be seen with OS X. Rather than force developers to use the (IMO far superior) OpenStep frameworks, they sank valuable time and resources into updating crufty old MacOS APIs.)

Because of this, we are seeing fewer operating systems in the marketplace, and I think we're the poorer for it.

Anyway, even with languages - don't you think he has a point? When I started using computers scientists used Fortran and people I knew in industry used C and we both used Perl to fasten things together. If the only advance in that time is Fortran 90 and C++/Java is that comparable progress to moving from Vaxen to Beowulf clusters?

Yes there may be more progress in "Academic" languages, but one of his points was the lack of cross-talk...

I think that as in other sciences paradigms shifts and great innovations are rare, the rest of the time we witness constant refinement.

When it come to languages, I think most of the great ideas are quite old (these include the idea of HLLs, routines, standardized looping constructs etc.).

Some more recent, but still fairly old, ideas include HM type inference and functional programming a la ML (FP itself is not so new; see the discussion of APL for example).

It is amazing to read the Lambda the Ultimate papers and see how many of the cutting edge topics (continuations etc.) are already there.

Maybe it should be ponted out that the great advances in programing languages (like OOP and abstraction facilites) were the result of thinking about programming paradigms in general (what might also be called SE). They didn't come from 'pure' language research.

Multics has been much more influential, mostly through its estranged child Unix. For example, the shell came from Multics. Programming a large software project in a high-level language. I know there is more, I just can't think of any off the top of my head.

For more info, you can look at http://www.multicians.org/

But all in all, when it comes the the topic of programmability (which Pike emphesizes) I don't remember huge advances. Maybe beacuse what we have is the Best There Is?!