innocence

i recently discovered audioscrobbler, which seems quite neat, but unfortunately it doesn't support shoutcast, so i suspect its estimates of my musical preference will be rather skewed :-(

the top image is from Motif, the standard Unix toolkit. it shows two selected checkboxes and one unselected. the bottom image is from GTK+; it shows an unseleted checkbox.

see the problem here?

i hate gtk.

it took me 3 hours to find the bug in the following function:

static struct http_client *
new_client(e)
	struct fde *e;
{
struct	http_client	*cl;

	if ((cl = wmalloc(sizeof(struct http_header))) == NULL) {
		fputs("out of memory\n", stderr);
		abort();
	}

	memset(cl, 0, sizeof(*cl));
	cl->cl_fde = e;
	return cl;
}

can you see it? damn, that's i'm stupid.

so i just committed Solaris ports support for ircd-ratbox. it's yet another IO multiplexer in the same vein as kqueue or epoll, but much simpler.

  int i = socket(...);
  int p = port_create();
  port_event_t ev;
  port_associate(p, PORT_SOURCE_FD, i, POLLRDNORM, NULL);
  while (port_get(p, &ev;, timeout) != -1) {
    /* handle event... */
  }

for something inspired by SystemV, it's remarkable lightweight and flexible (it support timer events and so on as well, and can even do thread synchronisation).

unfortunately, of course, it's incompatible with every other IO multiplexor in existence. wouldn't it be nice if the next version of POSIX or XPG had a nice standard version combining the useful parts of all of them?

(also wondering whether i should create another journal to avoid boring people to death with tech stuff..)

encarta becomes a wiki? (via jwales_rss).

april fool's would be a lot more interesting if people did something that was actually believable. "wikipedia bought by britannica" is such an old joke it's painful watching people try it...

unix may have come a long way in terms of useability recently, but there are still a few remaining problems.

for the last few days i've been trying to work out why anti-aliased fonts weren't working with KDE. the option to enable it was there, but upon selecting it, exiting and then opening it again, it'd magically revert to unselected. with no indication at all of why it was happening. the problem turned out to be that my ~/.qt was owned by root, so it couldn't write the configuration file. and instead of, you know, giving an error message, it just failed silently. helpful.

however, i now have a desktop i can look at without going blind. yay.

word of the day: Googlesphere.

Catholic code will ban schools from hiring gay teachers. (via bekithewitch)

Tim made an interesting observation earlier. when the site's mostly working, but slow, or there are some bugs, everyone sits in #mediawiki complaining about the minor problems. but when we do something extremely stupid and the site is broken for 12 hours, everyone says "thank you" for 'fixing' it.

where is the logic here?

illustrated process for converting coke into crack, courtesy... US Department of Justice (!). (via fross)

find the shortest path between any two Wikipedia articles. (okay, this is really old, but i guess most people reading this haven't seen it.)

currently having an interesting discussion with Sj, which may result in some interesting Wikipedia-related things happening in the future. i'm also finding some Wikimedia people really grating recently :-( not sure if they've become more annoying, or i just haven't noticed before.

seeing all the hacks brad comes up with kinda makes me feel that i should start doing something useful - i'm sure there's a load of neat wikipedia things waiting to be done.

however, in lieu of doing anything new, here's something i made earlier: Wikipedia recentchanges in real time (Firefox only).

this lj drama generator meme finally forced me to implement the s2 auto-meme-cut-tagger from s2howto. i think this is one of the most useful things i've ever seen on lj :-)

racist murderer gets 5 years in jail.

i wonder about the motives of some people that contribute to Wikipedia.

every so often, someone brings up the suggestion that the content should be put in the public domain, rather than licensed under the GFDL. someone then invariably says, "but then evil corporations will steal our work and sell it". i'm not entirely sure why this is a problem. what are we writing for? i think one of the most important aspects of Wikipedia is the potential to distribute knowledge, for free, to everyone. it sounds like a cliché, but Wikipedia is the first really successful attempt at doing this, ever—both because of its open nature, and the "power of the intarweb"(tm) which has only recently come into existence.

in my opinion, we should be doing everything possible to help this happen, and that means making it as easy as possible to distribute what we create. and that means public domain - or else copyrighted but completely unrestricted use. when you have to credit everyone who every edited an article, or work out how to define the 5 most important contributors, it becomes a pain in the neck to redistribute it. we make up for that by letting people off with 'this article comes from Wikipedia', but that's not technically correct.

i understand the next version of the GFDL might help with this, but what are we trying to do? — why does it matter to us whether people are required to state that the content is from Wikipedia? the content will be around long after the people that wrote it are dead and buried, the corporations who "stole" it have gone out of business, and most likely after Wikipedia itself has ceased to exist. what good is a license then? it seems very short sighted to make such a fuss about licensing when it's really irrelevant.

the google hosting thing made it to slashdot. must be a slow news day if they're posting about this before any of the details are even known.

i'm just waiting until google buy yahoo and wikipedia and let people edit each other's journal entries.

XEmacs S2 font-lock. very prelimary, in that it sort of works, but no doubt has quite a few bugs - i know nearly nothing about emacs...

( example screenshot behind cut )

i'm trying to make my s2 style be valid xhtml. so far i've mostly succeeded, except for one particular lj issue. it seems that when you have a long "word" (which is defined as anything between spaces, so long URLs etc, count), it inserts a <wbr /> tag in the middle. i'm not sure what this tag is, but it's not standard html. i submitted a support request about it, and apparently it's not possible to disable this. slightly annoying.

ursamajor had an even more annoying problem with this. apparently some RSS syndicators will try to auto-link URLs, like lj does. well, that's fine... except when they're broken and do something like: (URLs wrapped for appearence sake)

  <a href="http://www.example.com/">
    <a href="http://www.example.com>link text</a></a>

because lj already <a>-ified the link. as it happens, that also tends to work fine. however, once you add the <wbr />, it all falls down:

<a href="http://www.livejournal.com/support/see_request.bml?id=414126"
  ><a href="http://www.livejournal.com/support/see_<wbr">http://www.livejournal.com/ 
  support/see_<wbr</a> />request.bml?id=414126</a>

oh dear. i wish people wouldn't try to be too clever for their own good so much...

following on my abigailb and myself's previous research into google counts for "hehe" and "haha", here's a new graph showing both for a slightly larger range.
( pretty graph )