Older blog entries for chalst (starting at number 177)

ncm on Academic Computer Science IV
ncm: Thanks. Most interesting to hear, besides lurid anecdotes, would be the names of some of the journals involved.

I can think of countless reasons why someone might prefer a second-tier posting to an elite posting: possible to get spouse a job, more freedom and resources when creating your department, better relationship to the university hierarchy, able to see yourself working better with the rest of faculty, and, of course, climate. Reputation of the university is unlikely to be the first consideration of all that many people when weighing up rival offers.

Why Purdue, though? I can see Princeton, though MIT, Stanford & Carnegie Mellon appear to be more attractive.

ncm on Academic Computer Science III
ncm writes: You're refuting Bjarne's assertion about his academic colleagues' attitude toward industrial programming by pointing out that Bjarne is himself an academic?

Well, no, I'm not refuting anything, but perhaps my skepticism needs more explanation. When you wrote: journal referees insist the industrial programming problems his students address are beneath notice, I assume that's Stroustrop's account. There are two interpretations of interest that I might place on this:

1. Stroustrup is quite right: these journal's have effective biases that mean they do not take the actual needs of industrial programming seriously; or
2. Stroustrup is not rendering a fair account of what happened. In particular, I note that Stroustrup has been free of normal academic publishing pressures for something like three decades: first because at Bell Labs he didn't have to publish, and now because as one of the few, true Big Swinging Dicks of language design he has a "Get out of Peer Review free" card, resulting in the situation where he can publish, but is not a very useful source of advice to his students.

   Academic publishing is a nasty business, often unfair, and often needlessly so. Most academics have to learn the shape of the forces opposing recognition of their ideas though the hard way and through learning from the advice of their colleagues. But the good news is that scholars can turn good ideas into publications if they stick at it, pretty much regardless of how the land lies. Perhaps Stroustrup has forgotten that?

In either case, the story is interesting and I want to know more. Since a plain request for details of your source failed, now I plead: how, wonderful Nathan, source of deep truths and amusing anecdotes about C++, did you learn of this story?

ncm on Academic Computer Science II
ncm writes: Um, from Bjarne Stroustrup? Maybe you would consider him biased, but journal referees insist the industrial programming problems his students address are beneath notice. Guess where the referees come from. (Hint: rhymes with "septicemia".).

Biased? Well, no not exactly, more incomprehensible. He is an academic computer scientist, prof at Texas, PhD from Cambridge, long history in one of the most academic of the industrial research boilers. It doesn't sound to me as if his remarks can be taken at face value.

Is there a particular quote of his you have in mind?

ncm on Academic Compuetr Science
ncm writes: Academics express open contempt for industrial programming and its needs.

Nonsense! Where on earth did you get such a silly idea?

Recentlog
ncm replies to redi and expresses doubt that academia could come up with a decent successor to C++. Why not? C++ was designed in an environment (early 80s Bell Labs) that more resembles academic research institutes than normal industrial R&D. It is hard to get funding to develop new languages, though.

Citation Plagiarism and Wikipedia
Geoffrey Pullum writes about a scholarly phenomenon some call "citation plagiarism", summarising the firey confrontation between Alan Dershowitz & Normal Finkelstein, and discussing whether citation plagiarism is, in fact, any sort of plagiarism at all.

I don't really have anything to add right now to what he said, but I observe that if it is a sort of plagiarism, then undergraduate authoring guides that forbid referencing Wikipedia as a source, will often be forcing students to plagiarise. I've long thought such guides are contrary to the scholarly standards that universities should be upholding, now I see a sharp-edged argument to deploy...

Postscript: While I'm linking to Language Log & posts on copyright, Roger Shuy has some interesting comments on another post by Pullum on copyright, and on alluding to a Dilbert cartoon.

Bram's case against the Axiom of Choice
Bram Cohen wrote: The axiom of choice is presented as 'obviously' true, an axiom which can be simply thrown on the pile in the interests of convenience, without raising any notable philosophical hackles.

...which is a rather strange thing to say, given how it is one of the foremost hackles-raising topic in the philosophy of mathematics. In fact, set theorists don't claim that AC is obviously true of any theory of sets, but instead they say there is a class of mathematical concpetions of set on which the axiom of choice (AC) comes out true, most importantly the cumulative hierarchy. There are very natural axiomatisations of set theory, particularly Fraenkel- Mostowski set theory, which validate the Zermelo-Fraenkel (ZF) axioms but not ZFC (ZF+AC).

The same game can be played with other axioms: if you think the sets of ZFC are "too large", you can drop the power set axiom and choose something like Kripke-Platek set theory, with no sets of cardinality greater than the set of real numbers. If you don't like the foundation axiom, you can choose to replace it by Aczel's AFA axiom, a set theory which is based on bisimulation, and so is useful for reasoning about event structures.

Bram goes on to discuss a puzzle, via Lance Fortnow, involving probabalistic reasoning over infinite sets. These sort of puzzles are a little like Zeno's paradox: as Bram says, the problem is trying to extend finitistic reasoning to the infinite case. However, AC is not to blame here: the problem is trying to naively extend the reasoning about conditional expectations to the infinite case: there is no right way to do this. In particular, observe that the 'strategy' is nonrecursive, and indeed even with the benefit of an oracle, cannot be applied in a finite number of steps. Observe further, that the equivalence classes are not going to be measurable.

Postscript: Changed title, fixed discussion of Bram's problem
A commentor in the thread to Bram's post has nailed it:

Yup, nonmeasurable sets.

Proof: for each of the countably many finite subsets A of N, consider the function fA that flips the colours of the hats of people in A. This function is obviously measurable (i.e., inverse images of measurable sets are measurable), and indeed measure-preserving and bijective. Now let S be your set of representatives. Then the fA(S), as A runs over the finite subsets of N, are disjoint, and their union is all of 2N. If any of them is measurable then they all are and all have the same measure (because fA is measure- preserving). But then by countable additivity the measure of the whole space is either 0 (if |S|=0) or infinite (if |S|>0), but in fact the measure of the whole space is 1, contradiction.

Summary of proof: consistently flipping any finite set of bits turns our set of representatives into another set that's "obviously the same size"; there are countably many of these, and they partition 2N, so whatever the probability p of having no wrong guesses, we have to have infinity * p = 1, which is impossible.

I expect this can be tweaked to yield a proof that the event "at least n errors" is non-measurable for each n -- i.e., that the probability of getting at least n errors is undefined -- but I'm too lazy to try. :-)

Recentlog
wingo wrote: Using C to interact with MzScheme: extreme badness.
What's the problem? Is it much more complex than guile? There didn't seem to be much there over and above the usual GC boundary ugliness, and they do give you the choice of precise vs. conservative.

argp: You probably would have enough certs to get at least Apprentice rating, if only you hadn't made so many certs. In your shoes, I would prune that list ruthlessly.

A Powerful New Weapon
Luke Gorrie wrote:

The year is 2007. Java and SOAP applications have taken over the internet. The default response to any HTTP request is "500 Internal server error." You, the fearless programmer, must devise a way to trick each server into not crashing on your requests, without the benefit of error messages, logs, or standard- compliance. XML whitespace tweaking? Tag reordering? MIME header reformatting? ROT13 encoding? Nobody hears your screams.

Your only chance is a powerful new weapon: http-twiddle.el.

The Alligator Calculus
Bret Victor has put together a visual representation of the lambda calculus, Alligator Eggs. It expresses the idea of variable scope using the idea of family relationships (and also colour), and so is meant to render the lambda calculus graspable by fairly young children. Carlin, my eldest daughter, is very much into family relationships, colours and crocodiles, but at 3 years, I don't think that she's really grasped rules well enough to get what's going on here.

An easy puzzle: find an alligator family whose normal form needs more than six colours. And a harder puzzle: find a family that you can show to be the smallest such family.

Via Phil Wadler.