Slashdot discovers Lisp

Very true. I'd guess that 90% of pg's audience for A Plan for Spam (which led on to the popularity of his later essays) came from Slashdot. It's where I first heard about pg, and Trevor and probably rtm for that matter. Lisp has been discussed for years on Slashdot.

Also, folks here may not realize it, but Slashdot still wields a mighty traffic hammer; it is not some irrelevant also-ran that redditors/diggers/HN readers might think. At least one of our websites has appeared on TechCrunch, Reddit, Hacker News, and Digg in some context (never a major story, but usually getting some votes and some discussion, and occasionally making the front page)...one time my co-founder mentioned something we were working on in a comment on Slashdot. It was the highest traffic day we've ever had (to this day, and our traffic has increased steadily since launch, so every few weeks we have a new peak; but always less than Slashdot day). Our traffic is pretty good, in general, for the kind of business we run. But Slashdot is the motherlode. I speak of it fondly, still.

Anyway, for a site with a user base that large and diverse, Slashdot isn't as entirely idiotic as it could be. It's certainly no worse than Digg, and only slightly worse than reddit. And 10+ years ago when it was the size of Hacker News, it was roughly equally interesting and smart, though its focus was somewhat different (less business, more Open Source). The signal to noise is too low for me to spend much time there, but it's silly to imagine that Hacker News would remain as good as it is with a user count in the millions, like slashdot.

I remember milk coming out my nose when I read this comment--I guess it would be four years ago now: http://books.slashdot.org/comments.pl?sid=147701&cid=123...

I think it runs deeper than simply the assumption of bad faith. When growing up children go through phases of conforming to their parents and then distancing from them, and also conforming to their peers and then distancing from (most of) them as well. This seems to be a part of the personality formation process - at some point people feel compelled to disagree in the most confrontational form (and even demean their opponents) not to look for truth, but to define their own self. They also often engage in group identity defining behaviors, such as rallies about persecution, "real men/hackers/programmers never/always do X" posturing etc. Also voting on agreement/disagreement axis (instead of useful/wasteful axis) is one of those. Incidentally, 18 yo is the typical drafting age for many draft armies, this is when "us vs them" mentality can be readily utilized.

What's remarkable is that with the advent of the Internet we have places where this process never ends. Rather than growing up, slashdot community is in perpetual adolescent disagreement mode. The most mature members are leaving and going elsewhere, to be supplanted by the newcomers.

I'm not necessarily right, but this narrative helps me make sense of what's going on and provides a perspective on the community dynamics.

The two questions that interest me now are:

1. Where are new HN members coming from?

2. Where are they going to when they leave?

Common Lisp is sometimes frighteningly more bare-metal than C. I spent all last night moving stuff from kernel-space, into user space, messing with them in Lisp and sending them back up stream to the networking stack. By hand :-)

In C, even though you work with low-level memory regions and system primitives, you're not interacting with them. By the time you see side-effects the process has long exited. In Lisp you can have open descriptors floating around in your shell where you can echo and read stuff to them; you can share a working buffer with a unix system call or signal, and see it get filled, changed and emptied! To realize that you can layer higher-level conceptual tools over these primitives, including the entire CLOS metaobject protocol, is something that's utterly gorgeous.

See what iolib does to Unix, specially see the multiplexing framework; that's right, first class CLOS classes as wrappers for poll, select, epoll and kselect; your choice. Writing specializing methods for system calls is something no C programmer could ever imagine, much less enjoy ;-)

He didn't benchmark it against nginx. Per the linked presentation, all he actually benchmarked it against was mongrel and lighty w/fastcgi for php.

Sounds like something the Alioth performance game was meant for. I'd point him towards chameneos-redux; the current haskell implementation (http://shootout.alioth.debian.org/u32q/benchmark.php?test=ch...) is taking 3.73s to run, while the C champion (http://shootout.alioth.debian.org/u32q/benchmark.php?test=ch...) is taking 12.29s. The current C implementation is also small, so I would imagine that it qualifies as a "relatively simple task".

Why?

He's correct. Given enough time, he probably can recreate any lisp program with a faster C program. A good assembly language person can do better still.

However, if we're looking at the total time to running said program 1000 or even 1 million times, he's likely to lose.

While he's recreating a given program, the lisp programmer has produced N more programs. He'll never keep up.