Musings

The Bogdanov brothers insist that they are not committing a hoax. They’ve written to Aaron Bergman, insisting that they are on the up & up.

I’ve read their Annals of Physics paper, and am convinced that it is well-executed nonsense.

The editors of Social Text claimed that they were taken in by Sokal’s phony article because they thought that he was sincere. Sokal’s rejoinder was that his sincerity was irrelevant to whether the contents of his paper were sense or nonsense.

Here, too, the “sincerity” of the Bogdanov brothers is surely a non-issue.

I’ve been reading Dijkgraaf, Gukov, Kazakov and Vafa. They point out the obvious fact that the Vandermonde determinant (which arises from fixing the overall U(N) symmetry of the integral) that appears in the Matrix Integral can be represented using Fadeev-Popov “ghosts”.

So far, not too surprising. What had not occurred to me before they mentioned it, is that this leads to a much simpler set of Feynman rules for the perturbative evaluation of the Matrix Integral. It’s obvious, in retrospect, but leads to some very nice calculations.

I think I really scrooched the pooch on that one.

Converting this blog to .xml with MathML extensions proved even worse for accessibility than I imagined. Of course it rendered beautifully in Mozilla. But Internet Explorer refused to render the page at all, and OmniWeb decided to download the source, rather than attempting to render it in the browser.

I guess we’re not there yet, in terms of MathML. (Heck, maybe we’re not there yet in terms of XHTML.) I don’t know what that means for the future of using weblogs to comment on the archives. In the short term, it seems more than a little inadequate.

Anyway, I will leave a copy of the current index in MathML-enabled format, so that the brave among you can see whether your browser is manly enough to render it.

In my previous entry, I gave an example of MathML in action. Unfortunately, Mozilla doesn’t recognize that a file contains MathML unless it ends in .xml. A .html file just won’t cut it. It took me a looong time to figure out why I couldn’t see the syntactically-correct MathML rendered in my browser.

Eventually, I just changed all my templates to spit out filenames with extensions .xml, and all was good. But then I had to fool web surfers into downloading files with .xml extensions when they had gotten used to the .html counterparts.

Mod_rewrite to the rescue! I created the following .htaccess file for my blog directory,

# based on example in 
# http://httpd.apache.org/docs-2.0/misc/rewriteguide.html
RewriteEngine on
RewriteBase   /~distler/blog/
#   parse out basename, but remember the fact
RewriteRule   ^(.*)\.html$              $1      [C,E=WasHTML:yes]
#   rewrite to document.xml if exists
RewriteCond   %{REQUEST_FILENAME}.xml -f
RewriteRule   ^(.*)$ $1.xml                   [S=1]
#   else reverse the previous basename cutout
RewriteCond   %{ENV:WasHTML}            ^yes$
RewriteRule   ^(.*)$ $1.html

deleted all the vestigial .html files, and I was good-to-go.

“With enough eyeballs, all bugs are shallow.”
— Linus Torvalds

In my previous comments about the Bogdanov hoax, I alluded to this oft-quoted maxim from the world of open-source software, implying that analogous benefits flowed from the use of the eprint archives to disseminate research results in physics.

But is the analogy really correct? What if there really weren’t a mechanism for the thousands of readers of the source code to report back on the bugs that they have found (or to suggest improvements), except perhaps by releasing software packages of their own? Would “Torvald’s Law”, as it is sometimes known, still hold? Would we really still receive full benefit of those thousands of eyeballs?

Or, to put it differently, does hep-th need the analogue of Bugzilla?

Well, Bugzilla is probably not quite the right model, but I do hope that perhaps the weblog might provide such a model. While this idea has occurred to others before, three developments make me sanguine that perhaps now might be the time for it to take off:

1. The development of quality weblogging software like Movable Type eases the burden by automating most of the content-management.
2. The Trackback feature of Movable Type provides an automated way to link back to another site which references a given article. Without this feature, it would be hard to discover what others might have said by way of followup on a given topic. Since there is an Open Specification for Trackback, it can be incorporated into other software, perhaps someday including the archives themselves (“Click here to see commentaries on this paper.”).
3. MathML 2.0. As you can see from previous posts to this weblog, math support in HTML is crude, at best. But MathML is now standardized, and at least one browser (Mozilla) is sufficiently compliant with all the latest standards to implement it.

This last point is, I think, the diciest. If we wish to use this venue for discussing physics, we need a nice way to do math. There are automatic converters, say, from LaTeX to MathML. The difficulty is on the browser end. If we start including snippets like

${\displaystyle \coprod _{i=0}^{\infty }{x}_i}$

in our posts, they will be inaccessible to those with “lesser” browsers like Internet Explorer. This is not a matter of using “proprietary” features of some particular browser. I validate all these pages to ensure that they are perfectly conformant with the published XHTML standards. Its just that not everyone has (yet) implemented those standards.

Or maybe a little elitism is just what we need anyway. What do y’all think?

Stung by the Sokal Hoax, two French brothers, Igor and Grichka Bogdanov, have made a stab at righting the scales. They managed to obtain PhDs from the Université de Bourgogne in Dijon and publish 4 papers filled with nonsense:

KMS Space-Time at the Planck Scale, G. Bogdanov, I. Bogdanov , Nuovo Cim. 117B (2002) 417-424.
The KMS State of Space-Time at the Planck Scale, I. Bogdanov , Chin.J.Phys. 40 (2002) 149-158.
Space-Time Metric and the KMS Condition at the Planck Scale, G. Bogdanov, I. Bogdanov, Annals Phys.296 (2002) 90-97.
Topological Field Theory of the Initial Singularity of Space-Time, G. Bogdanov, I. Bogdanov, Class.Quant.Grav. 18 (2001) 4341-4372.

What this says about the French PhD system or refereeing in the journals (and some journals they chose! Nuovo Cimento? The Chinese Journal of Physics? Gimme a break!) is anyone’s guess. But the most curious aspect of the affair was that they never even bothered to submit their masterpieces to the eprint archives. Which is to say that no one (and I do mean not a single person) actually read this stuff before the story broke…

For those unfamiliar with the field, let me explain what that means. The abstracts of new papers submitted to hep-th are read daily by thousands of physicists. If the abstract sounds interesting, hundreds will download and read it. Feedback (positive or negative) comes swiftly and copiously. If (like 99.9% of all scientists), you want to get your work noticed and read, you send it to the archives.

If, on the other hand, you want your work to “fly under the radar” and make it into a journal, having been read only by one other person (the overworked referee, and, in this case, perhaps not even by him), then you studiously avoid sending your work to the archives. Since no one in our field reads the journals anymore (why bother, when the archives are so much more convenient?), no one will be the wiser.

For more discussion of the unfolding “scandal”, see this Usenet thread.

The plot thickens: The Bogdanov brothers are not listed among the current or recent PhD students of the Lab. Have their names been pulled from the website, or is this all a put-up job?

Apparently, the Bogdanov brothers have been at this game for a while. I’ve been pulling up reviews on the web of their 1991 book, “Dieu et la Science” (God and Science), written with Jean Guitton of the Académie Française. From the excerpts, it appears to be written as an extended “interview”, with Guitton as the interviewer, and the Bogdanov brothers responding (with what sounds to me to be mostly bullshit). But … and here’s the interesting bit … all the reviews I pulled up from Google credit the Bogdanov brothers as having PhDs in theoretical physics and astrophysics (when they wrote the book, in 1991!).

I suppose it should come as no surprise that blogger.com was hacked today.

Most interesting, I thought, was the relatively warm praise this piece had for the security of Movable Type compared to blogger.com. Those of you who have been reading these pages know about my grumblings about the security of MT. I can only imagine what that means about blogger.com.

Overall, I think it’s a good thing that a greater focus on security come to this genre of software. Pity if it comes at the expense of all those blogger.com users.

I’m in Cambridge, MA, but I thought I’d mention a couple of papers of recent interest.

David Berenstein had a nice paper today in which he investigates a couple of gauge theories with quantum-deformed moduli spaces from the matrix-model approach of Dijkgraaf-Vafa. Probably the most interesting example is an SU(2) gauge theory with three adjoint chiral multiplets with superpotential

W = g ( Tr( XYZ+ZYX)–2Tr(X+Y+Z) )

This is the worldvolume gauge theory on a bulk probe brane near the C³/Z₂ ×Z₂ orbifold with discrete torsion.

Another intriguing (“intriguing” is all I’m about to commit to right now) paper appeared yesterday by Yonatan Zunger.

In the context of matrix theory, one can reproduce various D-brane configurations by letting the X’s of the matrix theory (at finite N, they are NxN matrices) in the infinite N limit turn into operators which act as derivations on some algebra, A.

Actually, even in empty flat space , we can consider X⁰ = d/dt + A₀ as a tenth “X”, which acts as a derivation on the algebra A= C(R)×M_N. In flat space, the rest of the X’s (acting by commutator, [X,.]) are “inner” derivations of this algebra. In matrix string theory, two derivations are “outer”, and eight are “inner”.

His general proposal is to consider an arbitrary algebra, A, with some number of inner and some number of outer derivations. This he proposes as a generalized D-brane configuration in the matrix realization. It is characterized by k (k >2) classes in the Hochschild Cohomology H¹(A) (essentially, the number of X’s which are outer derivations), and a class in the algebraic K-theory K₀(A), which gives the “charge” of the D-brane configuration.

Anyway, time to get myself some dinner. More later …

I don’t normally post on political topics. That’s not because I’m not passionate about politics, but because there are already enough blogs out there filling that particular ecological niche.

But this weekend, the Sunday New York Times Magazine carried Paul Krugman’s article on the rise of Income Inequality in America. It’s an important topic, because it serves as a backdrop for almost every other item of debate in the domestic political agenda – from taxation, to healthcare reform, to campaign finance.

In a nutshell, after a generation of relative income equality (the postwar era through the '70s), we have returned to the rates of income inequality (the top 1% receive 14% of the total after-tax income in the country) which characterized the Guilded Age.

And, despite what you may have heard, greater income inequality is not the necessary price of greater prosperity overall (“A rising tide raises all boats.” was the Reagan-era motto). It’s true that the US has the highest GDP per capita in the world. Sweden, the bête noire of conservatives, has a GDP per capita comparable to that of Mississippi, one of the poorest states in the Union. But that statistic misleads, because it is skewed by the overwhelming rates of income inequality here. Median family income in Sweden is far higher than in Mississippi – comparable to that of the US as a whole. And, of course, the Swedes achieve that while taking longer vacations, having longer life-expectancies, etc. The average American is not significantly better-off than his counterparts in other developed countries, and the admirable growth of the American economy “as a whole” has not trickled down to the average worker. Median Family Income (adjusted for inflation) rose only 0.5% per year from 1979 to 1997.

Anyway, it’s an important read. And if you want to delve deeper into the statistics, here’s a compendium of data sources for the article. You might also want to look at Income Inequality in the United States, 1913–1998 by Pikkety and Saez.

Yesterday, I gave a lecture to Samantha Krukowski’s Film class on Time. We talked about Relativity, the Twin Paradox, and all that fun stuff.

The excuse (rather than merely being a little field trip for them to see how the other half lives) was to try to see how ideas from modern physics might inform their artwork.

I don’t know whether they got something usable out of the two hours we spent together, but they sure were enthusiastic. As I discussed earlier, neat Science can sometimes inspire neat Art. So it’s worthwhile trying to bridge the gap between the Two Cultures.

And it can be a rollicking good time, too.

There’s a wonderful application of modular transformations and conformal mappings in Escher’s The Print Gallery. In it, one sees, through a row of arched windows, a young man looking at a painting of a harbour scene. The painting mushrooms out of its frame, and on the waterfront one sees the museum, with its row of arched windows.

What’s going on mathematically in the picture was figured out this past summer by some Dutch mathematicians. Their site is filled with all kinds of wonderful stuff, including alternate renderings of The Print Gallery using different modular transformations.

Consider an ordinary image on the plane. View it as a function of w, the complex coordinate on the plane. It is, of course, periodic under 2π rotations of the plane.

w ~ e^{2π i} w

A “Droste effect” picture (after the Dutch Chocolate company, whose boxes have a picture of a young woman holding a box which has a picture of a young woman …) has a second periodicity under constant rescalings

w ~ r w

for some r > 0 (r = 256 is the factor relevant to The Print Gallery).

This, as you know, defines a torus. And we can conformally map it via

z = log(w)

where now the periodicities are

z/(2πi) ~ z/(2πi) + 1 ~ z/(2πi) + τ

with

τ = i log(r)/(2π)

Next perform a modular transformation of the torus.

z’ = z/(cτ+d)
τ’ = (aτ+b)/(cτ+d)

where a,b,c,d are integers satisfying ad-bc = 1. Finally, conformally map back

w’ = e ^z’

The picture in the w’ plane is Escher’s picture, if the original “Droste” picture was the w plane.

You can see this all worked through in detail, with wonderful illustrations here.

Gotta keep up with the Jones’s!

The latest release of OpenSSH can be found here, along with its PGP signature. In light of recent events, be sure to verify the signature!

gpg --verify openssh-3.5p1.tar.gz.sig
tar xzf openssh-3.5p1.tar.gz
cd openssh-3.5p1
sh
LDFLAGS="-flat_namespace" ./configure \
    --with-tcp-wrappers \
    --without-rsh \
    --prefix=/usr \
    --without-lastlog \
    --mandir=/usr/share/man \
    --sysconfdir=/private/etc
make

(then, as root)

make install
kill `cat /private/var/run/sshd.pid`
/usr/sbin/sshd

Frank Ferrari has given a nice field-theoretic proof of (the simplest case of) the Dijkgraaf-Vafa proposal, using the methods of Intriligator-Leigh-Seiberg. The nifty thing is that the linearity of the effective superpotential, W_eff, in the coupling constants, g_p of tree-level superpotential W_tree= g_p Tr Φ^p/p (one of the main conclusions of ILS) is precisely equivalent to the statement of DV that only planar diagrams of the matrix integral contribute to W_eff.

Meanwhile, Becker and Constantin carry out some checks of Gukov’s conjecture about the form of the effective superpotential in String/M-theory compactifications with background fluxes.

Jaguar ships with xinetd 2.3.5. Apparently, there’s a DoS vulnerability in xinetd 2.3.4-2.3.7. It doesn’t sound that serious, and Apple will eventually include an update in one of their Security Updates.

But if you are running xinetd and are concerned, you can, in the meantime, download and install the latest version, xinetd-2.3.9. To install,

./configure --with-libwrap
make

(then, as root)

make install
mv /usr/sbin/xinetd  /usr/sbin/xinetd.apple
mv /usr/local/sbin/xinetd  /usr/sbin/xinetd
kill `cat /var/run/xinetd.pid`
/usr/sbin/xinetd

For those who missed it, our friend Paul Ginsparg (founder of arXiv.org and recent recipient of a MacArthur Award) was interviewed in the New York Times Sunday Magazine.

If the prose of this Switch story isn’t funny enough by itself, check out the photo of the woman. Apparently, in addition to being a “freelance writer”, she has a second career as model for stock photographs.

Postscript: The fun didn’t last long. Microsoft pulled this page from their web site within hours. Still, some people managed to snag a copy before it went “404”.

It’s only compatible with Movable Type, but Kung-Log is a much nicer Blog client than BlogApp.

Gotta get Jaguar on my iBook …

Strange paper today by Klebanov & Polyakov. They claim the Fradkin-Vassiliev theory (a theory with an infinite number of higher spin gauge fields in AdS space) is AdS/CFT dual to a large-N CFT on the boundary.

This is puzzling. The AdS/CFT correspondence requires (I thought) a diffeomorphism-invariant theory in the bulk. That is, either a gravitational theory, of a topological quantum field theory. The FV theory is neither.

What gives?

Near as I can figure, what we really require is a local gauge symmetry in the bulk which contains SO(d,2), the isometry group of AdS, as a subgroup. I think that’s what Fradkin & Vassiliev claim to have. The amazing thing (if true) is that this gauge group is not the diffeomorphism group.

Cowabunga!

A good dinner, and a little reflection made me realize that the obvious solution to the cleartext-password problem of Movable Type is simply to ssh-tunnel the control connection. Typing

ssh -L 8080:127.0.0.1:80 golem.ph.utexas.edu

gives me an encrypted tunnel to golem. Then I can point my browser to

 http://locahost:8080/cgi-bin/mt.cgi

to get the Movable Type configuration page. Voila! No more cleartext-password worries.

If you don’t (or can’t) go the cgiwrap/suexec route, the alternative “dumb” recommendation in the Movable Type setup instructions is to set the permissions on your blog directory to 777.

I suppose this is necessary if you don’t have administrator access on your web server. The UID under which the web server runs needs to have write-access to your blog directory. So you either use cgiwrap or suexec to grant the CGI-scripts write-access to all your files, or you make your blog directory writable by anyone (including the web server).

But if you do have administrator access, then there is obviously a more secure alternative: change the ownership of the blog directory to match the UID of the web server (under MacOSX, this is “www”).

Since some significant fraction of their users do run their own web servers, it behooves the MT people to explain this, rather than offering the other two, distinctly inferior, solutions.

If this is what their Installation instructions are like, it makes me wonder about the attention to security in the software itself.

I have to admit, though, that they have put together a very extensive blog management package, and made it freeware. So I shouldn’t complain …

Everyone and his uncle seems to be doing it, so why shouldn’t I? I’ve just installed Movable Type and have joined the legions keeping a weblog.

So far, I am unimpressed by the security architecture of the whole thing. Everything seems to be sent as cleartext over the network. And the setup instructions actually assert that using cgiwrap or suexec made the whole operation more secure, rather than less.

That’s so funny, I could cry.

Anyway, most of the setup/configuration takes place through a cluttered, but not completely incomprehensible web interface. I’m eager to try out BlogApp, at least for posting/editing the weblog entries.

Unfortunately, that will have to wait until I can upgrade my iBook to Jaguar. Which, in turn, will have to wait until those (ahem!) nice folks at P&L Systems update their spreadsheet, Mesa.app to work with Jaguar.