New feature for LAMBDA (poll)

Use of Cobaya. camb, CLASS, cosmomc, compilers, etc.
Post Reply

Would you use a user-submitted software archive on LAMBDA?

Yes, and I have files that I would submit to such a site.
5
42%
Yes, I'd find it useful but I don't have anything I'd submit.
7
58%
I'm not sure I see the point and/or things seem fine as they are.
0
No votes
 
Total votes: 12

Ben Gold
Posts: 81
Joined: September 25 2004
Affiliation: University of Minnesota
Contact:

New feature for LAMBDA (poll)

Post by Ben Gold » January 31 2006

I've been discussing with a few people the possibility of adding a place for user-submitted software and resources to LAMBDA (http://lambda.gsfc.nasa.gov/). There are already links to many major packages (CMBFAST, CAMB, RADPACK, HEALPix, CMBeasy, etc.) so I imagine it would be more useful for smaller bits of code, data files, third-party patches for the above packages and whatnot, and I'm open to suggestions about that and how to categorize things. The basic format would be something similar to arxiv.org, where once you register you can fill out a form with some basic info and upload files. The idea is to have a centralized archive of CMB-related software. There probably would NOT be discussion boards; this site seems to be serving pretty well for that.

Right now I'd like to gauge interest in such a thing, so I've made a poll. But if you have any suggestions, feel free to leave comments.

Ben Gold
Posts: 81
Joined: September 25 2004
Affiliation: University of Minnesota
Contact:

New feature for LAMBDA (poll)

Post by Ben Gold » January 31 2006

And just to clarify, here are a couple examples of things I have that I'd consider submitting:

1) I have a little package for integrating the equation of motion for a single scalar field during inflation and spitting out the resulting primordial power spectrum (which you could then input to CMBFAST or the like). It's something I wrote in my early days of grad school and it's quite simple (though maybe oddly written), but I wouldn't mind sharing it.

2) More recently I wrote a very crude MacOS X gui interface for SuperMongo. I think I must be the only one who makes plots this way, but again, I wouldn't mind sharing it (although it's not really CMB-related). This might force me to work out some of the bugs anyway.

I've also got several "add-on" routines I've written over the years to extend the capabilities of CMBFAST, but many of the features have since been independently added to the main version (and done better) so I don't know how much good it would be to actually submit them. But that sort of thing would be a very good candidate.

Anyway, that's at least -my- idea of the sort of stuff such a site could hold. I'd love to hear other ideas!

Sarah Bridle
Posts: 144
Joined: September 24 2004
Affiliation: University College London (UCL)
Contact:

New feature for LAMBDA (poll)

Post by Sarah Bridle » February 02 2006

I think this kind of code sharing makes a lot of sense, and having some sort of archive like this is a great idea.
In an ideal world then people would provide the codes they used to make plots for papers. Is a bit idealistic, but at least having an archive is a start.

It is just a pity that Lambda is restricted to CMB. I think it would be more useful to have something that covers all of cosmology... and/or astronomy ... and/or everything - in a similar way to how arxiv covers a range of subjects.

Ben Gold
Posts: 81
Joined: September 25 2004
Affiliation: University of Minnesota
Contact:

New feature for LAMBDA (poll)

Post by Ben Gold » February 03 2006

I agree, and maybe LAMBDA can provide some momentum for a more general effort. After all, arxiv.org started out mainly as a service for high energy theorists (and there's still some astronomy that doesn't show up there).

Anyway, LAMBDA is restricted to microwave background analysis but if you poke through it you find that can cover a surprising range of stuff (there are maps of HI regions, dust maps, radio sources... etc). And that's the "official" supported data. I can talk with people and see how broad they're willing to stretch things. I don't think anybody envisions something that'll be policed terribly harshly anyway (mostly because that would require a lot of work).

I also realize that I may have mistakenly overemphasized the extent to which this would be an archive of source code. It'd be nice to have that there (and that's what I myself might use more), but there's a lot more on LAMBDA than code. For example, there are a few "cleaned" full-sky maps up there (WMAP's ILC map, the LILC map, Tegmark's cleaned map), but if somebody has their own that they've made this would be a perfect place to make the map itself available. I don't know if this would affect anyone's response to the poll.

The poll still has pretty small statistics, but I'm seeing an unfortunate trend toward "I think it's great, I hope somebody else helps make it." Is anyone willing to post and explain what might make them more likely to contribute to such a thing? Now's your opportunity to shape how it might work.

Ben Gold
Posts: 81
Joined: September 25 2004
Affiliation: University of Minnesota
Contact:

New feature for LAMBDA (poll)

Post by Ben Gold » May 03 2006

Just a quick post: This project hasn't been forgotten. The LAMBDA computer people have been quite busy with the WMAP release and some important long-delayed hardware upgrades, but things should start moving soon.

So if you have any suggestions, now's the time! Better yet, you can start preparing stuff for submission. :)

Boud Roukema
Posts: 87
Joined: February 24 2005
Affiliation: Institute of Astronomy, Nicolaus Copernicus University
Contact:

licences: GPL, public domain, etc

Post by Boud Roukema » May 04 2006

A difference with arxiv is that arxiv avoided copyright problems at the beginning by arguing (at least, this is what I heard by word-of-mouth in the community) that preprints are not the final versions of articles, so there's no conflict with the academic journals' copyright claims. From what I remember early on, the recommendation was to have at least one figure missing from the arxiv version in order to avoid possible problems, but by now, all the astro journals have been forced to accept arxiv publication without any need for doing stuff like removing a figure.

Since most of us only publish our software on our websites, probably we would not have the same potential conflict with "publishers". On the other hand, i'm not totally sure it will be quite so "simple" with software - I think most cosmologists just want to use/write good code and exchange it freely with colleagues, but we do have to coexist with the existing commercial/legal/political world to some extent. If we publish stuff as "public domain", then we take the risk that later on it could be commercialised and microsoft will make M$ out of it while we remain humbly paid cosmologists. On the other hand, publishing stuff as GPL can require a bit of rewriting your code in order to avoid dependencies on GPL-incompatible code, and even with different free (as in free speech) software licences, some of these licences are mutually incompatible.

http://www.gnu.org/licenses/license-list.html

Because of the complexity of licensing questions, I think some cosmologists tend to write their own copyright text, with the best of intentions, but without realising that this will make their own or their colleagues' use of the code unnecessarily messy and complicated.

Anyway, my suggestion: - make sure there's something like radioboxes (boxes to tick in) so that when people upload software they are forced to state what sort of licence the code has - people can come back to cosmocoffee to discuss this, e.g. help code authors clean up/remove/fix bits which may be incompatible with the declared licence, etc.

Options for licences should IMHO have a link to

http://www.gnu.org/licenses/license-list.html

and include options like:
  • GPL
  • public domain
  • other - please specify
This list reflects my own bias in favour of the GPL :), so probably some other common choices should be added too.

We could possibly also have a default, e.g. say that codes are assumed to be licensed under (default licence) if no licence is explicitly stated.

Any other thoughts on this?

Ben Gold
Posts: 81
Joined: September 25 2004
Affiliation: University of Minnesota
Contact:

Re: licences: GPL, public domain, etc

Post by Ben Gold » May 04 2006

I suspect in cosmology most of us don't really have to worry about our software ever finding commercical use ;). But I think your suggestion is good: there should be a default license and some options available for alternatives. I should note that even under the GPL, though, most of the license is irrelevant if you just modify code for private use and never redistribute it, which is probably the most frequent thing for researchers to do.

It's also interesting to note that CMBFAST itself is more or less under the MIT license (aka X11, aka modified BSD, etc). CAMB doesn't even seem to contain any license statements other than those already in CMBFAST, and appears to be intended as public domain. So if Microsoft wants to put CMB code into the next version of Windows, they're already free to. We all know who wrote it, anyway.

Post Reply