[clean-list] Clean, applied domain, goals(benefits), means,and investment

Marco Kesseler m.wittebrood@mailbox.kun.nl
Thu, 17 Oct 2002 23:24:53 +0200


Valery wrote:
>Marco wrote:
>> The academic research groups simply do not have the resources.
>
>why should anybody bring any resources if there is no language capable 
>of making explicit mass parallelism? There should be clear goal and means 
>to reach this goal to obtain resources from investors.

It is relatively easy to come up with languages that express massive 
parallel computation. Both implicit or explicit. I understand that 
you are enthousiastic about massive parallellism, but in my humble 
opinion, I think you will find that it is very hard to get investors 
interested.

There used to exist a massively parallell machine called "The 
Connection Machine" from the Thinking Machines Corporation. They went 
bankrupt.

But there seems to be a lot of interest in quantumcomputing...

>There are key questions: 
>
>what are...
>
>1. ... domain for typical software written in Clean?
>
>2. ... goals (benefits) to be achieved using clean? 
>(e.g. we could have better <whatever> using Clean)
>
>3. ... means? Or why could we have it exactly with help of Clean?
>
>4.  languages-competitors? (what are the languages competing in the same 
>     domain and pursuing the same goals?)
>
>If clean pretends to be more then just educational language (why not?) those
>questions should be answered.
>
>I see a good reason and abilities to conquer this game setting up the most 
>interesting kernel for Clean. I speak about an implicit parallelism. 
>Then it is quite easy to answer the questions above and make the picture 
>clear for investors.

In my view, Clean already is more than just an educational language, 
and it is because it runs decently on mainstream architectures. 
Nonetheless, I agree that your questions do need answering. I have an 
additional question:

5. Why is parallellism a prerequisite for answering these questions?

(Investors will ask the same)

>> So what if it slows down Clean by 10 percent?
>
>One could implement a simple database several times faster in C/C++ then 
>in mainstream database-oriented languages. And who denies database 
>languages because of it? Just imaging that Clean could became the 
>only one language (even being 20% slower as it might be) but capable 
>of making mass parallelism.
>
>Clean with its implicit parallelism might become the only (or the very
>first) 
>motivation for a new generation of hardware (however that's for other 
>topic/thread/discussion)
>
>> The reality of today is that the Intel platform has become the main 
>> platform for Clean. 
>
>And? Clean already has a very good performance. 
>What do you imply here? Do you want to compete C/C++ or ?..
>
>> I do not see how one can see forking implemented 
>> in Clean without discussing the Intel hardware.

Clean already competes with imperative languages. Haskell does this 
as well. 

I did not intend to imply anything. Intel is the main platform for 
Clean. If you want to have "forking" implemented, it will have to 
happen on the Intel Pentium and the PowerPC, because there is no 
implementation for other processors.

>sure. lets fork a topic then: "Forking on Intel" :)

Sure.

>> You could, but who is going to give this new machine to you? 
>
>with current goals and means? -- probably no one.
>
>> It is more likely that the rest of the world is going to say 
>> "see, these functional languages will never be efficient enough".
>
>Why should functional languages play smb. else's games?

Because there is little else to play with. By the way, the Intel 
architecture is not somebody elses game if your language can play it 
well.

>However this is easier to discuss after 4 key questions above 
>become answered.
>
>> Apart from that, people probably want a machine that can run 
>> different kinds of languages well.
>
>What's the problem with Clean here? It is already quite good on our 
>classical machines.

This is not a problem with Clean. It is an additional reason why no 
one will come up with specialised hardware for running Clean.

regards,
Marco