Older blog entries for haruspex (starting at number 12)

Never forget.

nymia wrote,

We all know Java can do Write-Once-Run-Anywhere (WORA), but Javascript can do better than that. Actually, WORA is a little bit different because Javascript code is JIT'ed, also known as on-the-fly translation.

Isn't it the other way around? I haven't heard of any JIT JavaScript translators, but there are of course many Java ones. (Unless you're saying that JS is interpreted on the fly, which is true.) My hunch is there are few common uses of JavaScript where JIT translation would deliver much payoff.

One related difference between Java and JS is the executable format: Java is precompiled bytecodes, while JavaScript is always reparsed from source (anyone know any exceptions? Do modern JS interpreters tokenise for later?)

Not that there's anything wrong with JS. It's cool.

amars: the short answer is, throw away the low order bits, e.g. (rand()>>16) & 1

The construction of good random number generators is a decades-old subject now. It's a job for wizards. There are many pitfalls and quality failures; useless low order bits is just one of them. This problem with BSD rand() is fairly well known, and is often remarked upon in the source code and man pages.

The low bits of the numbers generated are not very random; use the middle bits. In particular the lowest bit alternates between 0 and 1.

If your system's RNG is flawed, you can just use a better one, e.g.:

Donald Knuth's RANARRAY

the Boost library

ThinAir (more references on that page, too)

Taygeta's bibliography

thomasvs: Why not use the ISP's SMTP server directly from your home machines via something like NAT? I think that would solve it.

This evening in Melbourne (Australia), a talk on "incrementally" implementing Agile/Extreme Programming principles, given by Steve Hayes and hosted by the local XP User's Group. A solid and enjoyable presentation by Steve, who speaks from extensive experience and obviously knows his stuff! and an excellent round table discussion. The PPT show might end up on his web site.

snim2: Much as I would like to see the SCO board behind bars, ESR doesn't speak for me either.

mrorganic: The idea of UNIX-as-literature has a lot of basis. But don't omit Multics from your Alpha-Omega.

Summary: Unix is about words. Certain kinds of people are drawn to using words over other forms of interaction. Words rule. :-)

• Oblomovka writes (and here) that the BBC will open its Archive of radio and TV programming to the Internet. This is extraordinary news. In a bleak global context, when every news item is either documenting the progress of modern Stalinism or the timeless duel between greed and freedom, this is a truly inspiring project, that promises to reveal one more glimmer of the Internet's potential. God bless the BBC for taking the plunge. (Also see /. thread)

• Is it not strangely fitting that a thread on alt.folklore.computers headed Dealing with complexity has at present reached the impressive total of 1956 posts. This may stretch Google's thread navigation mechanism to its limits... perhaps there are better ways of browsing threads. I know Geocrawler's is about the worst ever (i.e., no mechanism at all).

Jimbob: thanks for a breath of sanity and intelligence. With you all the way.

jlouis:

Not Intels CC. Commercial dependence is bad.

Uh, yeah, and the fact that it targets only ONE popular but uninteresting architecture. To enumerate gcc's CPU targets alone, I'd need all my fingers and toes... and then some.

It's taken gcc over 15 years to get this far (how many man years?) and a competitor has a lot of catching up to do. Expect a long haul to reach a clean build of e.g. the major gcc-built O/S and their port libraries, multiplied by all their supported platforms. Would it really be worth it, to finally reach similar quality and performance, but still be dealing with the same old language (which I have grown to love)? Try innovating instead (C# doesn't count).

Can "monopoly" mean anything when the product is free in every sense?

It's myopic to try to equate any commercial Wintel compiler with gcc. What gcc has done is openly solve the compiler problem, once and for all, so that great minds like yours can concentrate on other problems. (Thankyou RMS.)

How can one go further back in the "recent diary entries" list? I seem to recall someone posted a neat definition of hacking in the last couple of weeks and I would like to find it again...