Comments on: Welcome to CCC

By: drj

drj — Fri, 04 Dec 2009 08:56:59 +0000

[drat: didn't see Nick's reply before posting this. Sorry to sound like a repetitive bore]

@mathew: That’s an interesting point, but I don’t see how a formally verifiable language would contribute to any of the goals. Especially goal 1: “To produce clear climate science software”.

Ravenbrook in general like to stress the importance of clarity. For this project clarity is paramount. The clearer it is, the more people will look and understand it. And thereby hopefully achieve goal 3.

Did you have any particular language in mind?

By: drj

drj — Fri, 04 Dec 2009 08:53:00 +0000

@Richard That's excellent thanks. I see you've been filling in issues already. Thanks for that!

By: Nick.Barnes

Nick.Barnes — Thu, 03 Dec 2009 21:44:11 +0000

Paul: Yes, we do have access to a statistician. We’re not going to start looking at that until after we’ve finished the first part of the project (i.e. producing a clear working re-implementation of the existing algorithm).
I see our critical job as exposing the algorithms to public view and understanding. When it’s all in old Fortran, the algorithms are too obscure for anyone to understand.
So although I do have plenty to say about the particular algorithms in GISTEMP (and why I prefer the simpler alternative here: http://ds.data.jma.go.jp/tcc/tcc/products/gwp/temp/ann_wld.html ), that’s not a job for today.

By: Richard Hendricks

Richard Hendricks — Thu, 03 Dec 2009 21:32:50 +0000

Happy to contribute my time between Windows CE compiles. I’ll try and go through a typical “user experience” first and see what kinds of issues they would run in to and file tickets.

By: Paul

Paul — Thu, 03 Dec 2009 21:26:21 +0000

What are you planning to do about the statistics used? When reviewing the “Harry readme” (not for gis-temp, of course, but others have said similar things about gis-temp) I felt there were several issues – coding was one of them, but one of the others was the statistical validity of the things done to the data – do you have access to a statistician?

By: Nick.Barnes

Nick.Barnes — Thu, 03 Dec 2009 17:52:55 +0000

I did think of that, mathew, and as you probably recall I have a particular affinity for SML. But:
1. As far as I know there is no formal mathematics around GISTEMP, to which a proof could relate. Scientists basically don’t do formal mathematics (with some exceptions in physics);
2. Fewer people would understand an SML implementation;
3. Fewer people could contribute to an SML implementation;
4. (weaker) Fewer people could run an SML implementation;
5. Python comes with a bunch of stuff out of the box which is pretty useful (e.g. a BSD-DB interface).

If we want a formalism, we can certainly write theorems about a Python implementation.

By: mathew

mathew — Thu, 03 Dec 2009 16:24:10 +0000

I can’t help wondering if a language that was amenable to formal proofs of correctness wouldn’t be a better choice.

By: Nick.Barnes

Nick.Barnes — Thu, 03 Dec 2009 10:45:45 +0000

The GoogleCode project has an issue system, here. It's not very populated yet.

By: John Keyes

John Keyes — Wed, 02 Dec 2009 23:49:55 +0000

You’re very welcome. Looking forward to seeing this project developing further.

Is there a ticket/issue/bug tracker or a TODO list where people can see where help is needed?