Lambda the Ultimate

It's one thing to say that "academics should as far as possible see if they can do their work outside the ACM"; however, as far as access to papers goes, the ACM has a total monopoly on many of the papers in its digital library. For someone trying to come up to speed on the field, boycotting the ACM would be pretty counterproductive. Hence the grumbling, I suppose.

POPL sort of dominates the field.

Hmmm... Lots of researchers put lots of good papers online. Springer LNCS also has lots of nice stuff, and AFAIK everything older than two years is 100% accessible online.

There have been papers where citeseer showed me that nice not-in-database page, but then the abstracts didn't sound THAT great that I would resound my principles.

Most good stuff is available outside the ACM, so I think those guys should get off their high horse. If a researcher thinks he has to publish a paper exclusively in an elitist/not-open "company", then he deserves not to be read IMHO.

I really hope to see more grassroots events like the European Lisp Meeting 2005 taking over from ACM workshops too.

As a former US taxpayer it irks me that my taxes funded any research that I am not allowed to read.

As a self-employed programmer, I can't afford to subscribe to dead-tree journals, even though I could often use the information in my work.

I was thrilled about scholar.google.com but it's frustrating to use when some of the papers I find are inaccessible to me. I sent in a feature request asking for the ability to filter out subscription-required results.

My partial solution to these issues is to organize The Monad.Reader, a free online Haskell eZine.

As for grassroots events, I was part organizer of EuroHaskell in 2004 and we'll be doing it again this year.

We can fix these problems ourselves. If your favorite language doesn't have a free online magazine, start one! If you've already got one, write articles!


--Shae Erisson - ScannedInAvian.com

After you've found an interesting pay-per-download paper with Google Scholar, search again on plain Google. More often than not you'll find a freely available version on the author's homepage that for some reason wasn't catalogued on Google Scholar. I think Google Scholar is a little picky about what it considers to be good sources of scholarly material.

Amplifying what Shae said, there's a very important group out there, the "armchair computer scientist". These are the people who are sophisticated but not necessarily in-depth, and are interested in good theory, but don't have the time for a literature review or chasing citations. There's a lot of scope periodicals like The Monad.Reader which make this stuff accessible.

(And no, I'm not just saying this because I have an article in the first issue!)

It would be nice to get free access to the ACM archive. However, as it stands, I'm happy to pay their student fees and get full access. Believe it or not, the ACM is relatively well-priced in this regard - you can get full access (and a student membership) for around $42 a year (I forget how much the price is for a regular membership). Compare that to the IEEE's $35 PER MONTH, (on top of your regular membership fees), and 24 paper/month limit. The situation with the ACM, while not necessarily ideal, could be FAR worse.

Funny, the same idea came to me while reading the previous posts on the thread. I am not a lawyer (and I don't like the usual abbreviation either) but I suspect that given enough fire, err, lawyerpower this procedure can be presented as a mere translation of the original work, which I believe is covered by copyright laws. Or do translations have to be literal? It probably depends on the jurisdiction, but in general?

In CS, a lot of papers start, or end, or are copied, as technical reports so I don't think there really is a need to do that.

Even though it is in fact possible to find much of CS literature (well, at least of the past 10 to 15 years) online for free, it is quite a bit harder there than through e.g. portal.acm.org. And for older stuff you're still up shit creek. I've been doing some research into Algol 68, and I was damn glad the university has full portal.acm.org access...

I used to download and read a lot from Springer's LNCS.
It used to be free, no registration, nothing required, but articles were only online after two years from publication.

I just looked and it seems they moved to the ACM-like pay-for-article, with only the abstracts online.

Does anyone know what happened? Are all articles that are two years or older, still freely available somewhere?

They realised that they could get people to pay for the stuff?

Unfortunately, they use the metapress platform now, which is absolutely awful (for example, one cannot bookmark the pages that one views - one has to manually bookmark the "bookmarkable" URLs they display, not as links, either).

As marco comments, I don't think that any such effort is necessary in most cases, because most papers not available in their published form free online are available free, with identical text. However, emphatically, there is nothing preventing you from summarising another's work, commenting on it, even to the point of reproducing all results, as long as you don't try to pass it off as your own, or use any of the actual text. (IANAL)

This is done on lots of papers, thesis, etc. It's very common to people to create distill and re-write a lot of papers and call that a graduation thesis. Also, it very well accepted on the academia, some digests have even more citations than the target paper.

The only problem is that you may want to check what the original author said and that those digests are normally not freely available.

Remember, sicence is not only testing and proposing new hypotesis, but also doing that:
"* Acquire (legally) a paper from the ACM or some other pay-through-the-nose academic publisher
* Distill the knowledge in the paper to it's essense
* Re-write the paper, so that it contains the same essential knowledge, but with prose and explaination written from scratch.
* Cite the original paper in the bibliography (also include all works cited in the original paper)
* Publish the result on CiteSeer or some other free digital library.
* Openly acknowledge, in the introduction to the paper, that the above procedure was done.
" [father post]

I know very little about the economics of academic publishing, but...

* Authors don't get paid for papers--in many cases, they are required to assign copyright to the journal as a condition of publication.
* Peer-reviewers don't get paid
* Producing journals in dead tree format does incur costs that aren't incurred by online publishing (paper, ink, and postage mainly)
* Editors and such (the folks responsible for assembling the journal itself) do get paid, but how much?

So... where does all the money go?

Reed Elsevier, which has the largest market share in scientific publishing, has an operating margin of 34pc

(from business.telegraph)

Authors are often charged for publication. Staffing, promotion, and distribution, all cost, however, and subscriptions for the whole dead-tree journal are very low, typically.

The problem here is that academic publishing is bound-up with paper-based publications, which have a finite capacity to carry papers over unit time, and so incur backlogs of papers to publish. Sadly, no-one has yet sorted out a good way to replicate the peer-review process for a free online journal.

The advantage of paper-based publication is the redundant replication of data and distributed timestamping of publication that that generates.

LMCS might be of interest to LtU readers. From the purpose page:

Logical Methods in Computer Science is a fully refereed, open access, free, electronic journal. It welcomes papers on theoretical and practical areas in computer science involving logical methods, taken in a broad sense; some particular areas within its scope are listed below. Papers are refereed in the traditional way, with two or more referees per paper. Copyright is retained by the author.

The editorial board has many of the PL community bigwigs, Dana Scott, Benjamin Pierce, Philip Wadler, Mitch Wand, etc.

I was about to post this link, when marcin beat me to it. The rumors about the contents of the first issue are rather promising, and the editorial committee is simply stellar.

Seems to me that the programmer and academic cultures don't have much overlap. Not that academics don't program, but that programmers are more likely to look for skill improvement the java.sun.com website rather than lambda-the-ultimate. Commercial programmers don't see how research papers can be applied to their daily work.

What about a bridge between cultures?

Two ideas that come to mind are a wikipedia of applied research papers, and an academic counterpart to The Pragmatic Programmer. (I can only imagine what articles would be written by Peyton-Jones, Wadler, Steele, Cardelli)

In both cases the hook would be solutions to existing problems in existing languages. The line would point back to further reading on the same subject in SICP, CTM, HTDP, and other high calorie CS books. The sinker would be the elegance of those same solutions in Haskell, Scheme, Erlang, ...

As for releasing old papers, In the current business culture, companies have a death grip on their copyrights and patents (witness the recent EU patent vote). I think the best way to free information is for us to write up our own learning and studies.

Any other bridge building ideas? I'd like to find something worth putting in my time and money.

--Shae Erisson - ScannedInAvian.com

I think computer science lacks continuity because people are deprived of access to its history. A large and hungry audience awaits the good stuff on sites like this but it's not being allowed to flow.

Here, here! I completely agree. And there is a lot of good stuff. Most ideas in computer science were invented in quite a reasonable form in the 60s and 70s. I was amazed to find the concurrent Sieve program of CTM in a 1977 paper by Kahn and MacQueen. Silly me thought I had invented it first. (I am lucky to be sitting next to a really good CS library, which has works dating back to the 40s and 50s.) The only difference is that they didn't have an implementation, but the program is there!

How about mining those good old days for new ideas? I'm sure there are still plenty there, waiting for the first person who takes the time to dig them up again. If Google were to digitize them and make them searchable, what a windfall it would be.

I would love to see a compare and contrast of Oz's distribution subsystem with that of the OpenCroquet project, or more generally, with the principles set forth in David Reed's 1978 Ph.D. thesis.

I have answers, which is: divide up your work for publication into (Class i) the work that most furthers your involvement in a scientific conversation and (Class ii) the work that best represents what you are up to to outsiders.

For the first class avoid mass audience journals for that part of your work you care about most, but prefer in-house university publishers for printing, or be internet only; it is especially important with this kind of work to keep copyright with few restrictions on use. Look for journals whose audience most narrowly fits the profile of your ideal reader; these are the journals that will provide the best feedback. Ideally, these are the kind of journal in which publishing involves you in an ongoing conversation.

Use name publishers (OUP, MIT Press, De Gruyter are more prestigious than Elsevier) of name journals (JACM is the 800 pound gorilla here, but I'd also count I&C, MSCS, and APAL as this sort of journal; TCS has lost a lot of the cachet it used to have) for work in the second class. The aim of the exercise here is, firstly, to decorate your CV, second, to push up referenceability of your work, and third, to engage people who work outside of your core research interests. Avoiding copyright assignment and/or heavy restrictions on reuse would be nice, but it's not as important here.

Postscript (19 Mar 05): rewritten above, lots of new content.

There are journals that go by the subscription model. Many of them are not planning on changing their basic model any time soon. Some of them might want to gain visibility by working with LtU, but I am not sure there's any workable solution they can agree to.

Hence, my question. If you have ideas, let me know (privately, if you prefer).

I rewrote the parent of this node so that it is more accessible, and used the opportunity to add a bit more information.

One thing I have noticed happening more and more is researchers using their departments' technical report series in a more and more creative manner. The significance of tech reports today is quite different to what it was 10 years ago.