Re: [TCLCORE] Is Tcl bytecode engine is also *in the far future* going to LLVM?

[Donal K. F. <don...@ma...>]

Tomasz Kosiak wrote:
> But I do think that we may want to think about finding some funding
> for those improvements. Python has Google who pays 2 engineers full
> time salary. But we as Tcl community also have some companies
> involved. None to compare with Google, but some of us can provide some
> funds.
> 
> Especially I was thinking about what PyPy did. In European Union there
> are possibilities to use public funds - if you plan to release results
> of your research project to the public you can receive 75-100%
> financial compensation for the project.

Putting together a bid for funds from the EU is non-trivial! (I should 
know; they fund the projects that pay my salary...) But then again, I 
only really know the research vehicles within the DG-INFSOC system; the 
others are all ones that I've never examined in depth. But the amount of 
bureaucracy will be high.

> In fact you probably have to spend 50% of this on paying sponsoring
> university and for paperwork. But event if we will have to fund 10-25%
> of the research costs it is cheap. As there are some university stuff
> from Europe and some companies we probably can think more about it.

You're really thinking about a research project? Hmm. If I remember 
right, the rules for small companies in Framework 7 are fairly 
favourable (like universities, they'll be able to recover 75% of costs, 
but I don't know which cost models are admissible for SMEs) so it might 
be worthwhile examining what it requires. But we'll have to target 
either one of the later calls or FET-OPEN (effectively no close date, 
but exists only to allow funds to go to areas not predicted during the 
setting up of the calls). (Actually, Call 4 is still open but we stand 
*no* chance of making it as it closes on 1 April. I've heard of people 
writing a successful bid in two weeks, but not in two days! We're 
talking over 100 pages of stuff even with a small consortium and few 
workpackages.) Any bid *must* have partners from at least three member 
countries, and mixes of universities and companies seem to work well.

I've checked the rules/topics (downloadable from the following URL: 
ftp://ftp.cordis.europa.eu/pub/fp7/ict/docs/ict-wp-2009-10_en.pdf) and 
it seems that (assuming you're talking about Framework 7, ICT area, Call 
4) that we can't apply. This is because we're thinking about this about 
two to three months too late (the call closes on 1 April).

The best chance is FET Open. It might be a bit tricky (apart from 
everything else, we'd have to provide some evidence that we are not just 
some people who couldn't get their act together for one of the other 
calls!) but should be possible. And the process is much simpler to start 
out with.

The good news is that conditions for SMEs are indeed favourable.

Of course, if we were instead to focus on basing a project on using Tcl 
to do fancy virtualized services, we'd be in with a really good chance 
to do a good application for Call 5; the Objective ICT-2009.1.2 looks 
rather interesting, especially outcome b on Page 17. (It's also a 
relatively roomy objective; very unlikely to be crowded out by a single 
big project. This is Good.) While getting money to fund compiler 
development would be nice, getting funding is absolutely about the Art 
of the Possible...

> I know that this is a crazy idea but I would like probe it somehow? I
> know that Donal K Fellows and Gustaf Neumann seems to be European
> academics. Zoran Vasilijevic ARCHIWARE or my company DAC System are
> also located in Europe. Such a project should be lead be some kind of
> consortium. When I take a look at PyPy consortium it doesn't seems as
> made of large companies either - http://pypy.org/consortium.html. I
> think that outsourcing some work to US or Peru won't be a problem ;)

The US and Canada are curiously tricky in this regard - involving 
partners from there can be tough - but collaborating with most of the 
rest of the world will earn extra political marks. Promising to Open 
Source the results would help too, and shouldn't be a problem at all.

If we want to do this (targeting FET Open or Call 5) we should strive 
for it; I have a pretty good idea what a good application looks like, 
and I most certainly can get support from work for doing it. But *all* 
involved should be aware that this is a huge lot of work. After all, 
there's potentially a lot of money involved...

This takes us back to the questions:
  1) Are we really interested in doing this?
  2) What do we really want to do with it?
  3) Who will actually make sense to have involved?

If there's genuine interest in this, I'll get some closed mailing lists 
and other appropriate communication channels set up. Some arguments it 
is best to not have in public... :-)

(Also: Call 5 is not yet open, so it is possible - if unlikely - that 
the interesting parts might not be there or changed so that we don't 
want to apply. Can't say for sure until mid-July.)

Donal.