It would be nice to also show pictures near those places in the map flyouts, but you can't get a Snapzilla feed for a place. So I could additionally map recent pictures like the secondlife.com front page used to, or map globally recent pictures that are near the landmarks, but not the most recent pictures from that place. As soon as the picture gets old enough, it goes away, which would happen plenty often given how many more Snapzilla users there are than slurlmarker users.

It would also have been nice to do it completely client side, using del.icio.us' JSON API, but they only let you get a specific user's bookmarks with JSON. This view is built from everyone's links on the slurl tag, so I had to use Plagger for that.

I expect to add a helpful blurb to the page about how to get the slurlmarker HUD and post places with it, once the vendor is placed in world again.

• Arito Cotton suggested I shouldn't use LINK_ALL_OTHERS, but address the prim I wanted to talk to specifically. As the object had only one prim to begin with, I didn't think this would matter, but I'll give it a whirl.
• CrystalShard Foo suggested using llGetTime() instead of llGetWallclock(). The big yellow warning scared me off of llGetTime() before I read the "This timer should be used for performance-related uses only" at the end.
• fluffy suggested I should optimize my script by batching up the messages and sending them in single says. Unfortunately the really slow exchange is an A → B → A → B pattern. Really I need to speed it up by revising the protocol so it only needs one A → B exchange, which has nothing to do with llSay() versus linked messages.

So here's a new script:

default {
    touch_start(integer total_number) {
        integer i;

        llOwnerSay("starting timer");

        llResetTime();

        for (i = 0; i < 10000; i++) {
            llSay(1234, "hi this is my message");
        }

        float tSays = llGetAndResetTime();

        for (i = 0; i < 10000; i++) {
            llMessageLinked(1, 0, "hi this is my message", NULL_KEY);
        }

        float tLinks = llGetAndResetTime();

        for (i = 0; i < 10000; i++) {
            "hi this is my message";
        }

        float tNoop = llGetTime();

        llOwnerSay("Says:  " + (string) tSays);
        llOwnerSay("Links: " + (string) tLinks);
        llOwnerSay("Noop:  " + (string) tNoop);
    }
}

that yields new output:

[0:45] timer: starting timer

[0:56] timer: Says: 279.611328

[0:56] timer: Links: 276.838348

[0:56] timer: Noop: 108.153099

that shows pretty much the same thing.

default {
    touch_start(integer total_number) {
        float start = llGetWallclock();
        integer i;   
        for (i = 0; i < 10000; i++) {
            llSay(1234, "hi this is my message");
        }
        float end = llGetWallclock();

        llOwnerSay("10,000 says took " + (string) (end - start) + " secs");
    }
}

Object: 10,000 says took 229.000000 secs

default {
    touch_start(integer total_number) {
        float start = llGetWallclock();
        integer i;   
        for (i = 0; i < 10000; i++) {
            llMessageLinked(LINK_ALL_OTHERS, 0, "hi this is my message", NULL_KEY);
        }
        float end = llGetWallclock();

        llOwnerSay("10,000 linked messages took " + (string) (end - start) + " secs");
    }
}

Object: 10,000 linked messages took 237.000000 secs

I guess I need to optimize somehow else.

I'm using DarwinPorts to install the UNIXy stuff like Apache 2, Python 2.4, and MySQL 5. I looked at Fink, since that's the one I'd heard of long before I had the Mac, but as they say in their currently newest news entry:

Fink on the Intel platform is still considered "beta" quality, and a number of packages (particularly packages in the "unstable" tree) either do not compile, or compile but do not run. Work to improve this situation is ongoing.

Installing Apache was cake, but MySQL has been a headache. First I had to figure out I needed the +server "variant" to get the Launch Services script for actually running mysqld. (Getting that installed--making sure the +server build was "activated" and the previous wasn't--was another stolen night, but I'll cop to that.) Now I just get the Can't connect to local MySQL server through socket '/opt/local/var/run/mysql5/mysqld.sock' error as if it's not running, which I guess it isn't, despite there being some weird root process with "mysql5" in the name.

Oh, that's the problem from farther in the thread. I guess it'll have the same solution, too, as "set up the main database" is the next step in this "MAMP" guide. That guide's a lot more helpful, in fact; I'm sorry I only found it just now searching for that email thread.

Also, anti-props to everyone who thought it was OK to leave .turd_mysql5 files in the empty directories port created. I don't want your shit (npi) on my computer.