I'm still getting the NAT stuff set up on the xserve, and haven't even touched traffic shaping yet, but I was pretty amazed by how I managed to break things in the process of setting up a stateful firewall on OS X.

Apparently if you forget to allow connections via localhost, OS X sort of loses it. And I mean this to the extent that when logged in as admin, you no longer have an identity. su asks you who you are.

I guess I hadn't considered the ramifications of directory authentication, even when you're not doing remote stuff.

Here at the office we have a dual 2.3gig G5 XServe 1U as sort of the office file server. It's a shame, really, because literally all it's been doing has been just that: acting as network storage.

I've gotten fed up with the crappy little Netgear router that's running NAT, so I'm finally going to give the Xserve a little more responsibility and let it do NAT, DHCP and some traffic shaping.

The one thing that scares me a little, though, is that the server also has a couple printers connected to it and these seem to give it some periodic stability issues. For instance, I ssh'ed in just now, ran top and saw:

PID COMMAND      %CPU   TIME   #TH #PRTS
 82 atprintd    54.4%  120 hrs   1    15
 81 atprintd    54.2%  120 hrs   1    15
 80 atprintd    53.3%  120 hrs   1    15

So basically a dual-G5 is sitting at a load average of 3 running printer spools. Yuck. Googling for atprintd doesn't give me a whole lot on where to look for this one.

CUPS on Apple has in my experience been no more fun than CUPS on Linux. There I ditched CUPS for lprng.

From the network cabinet at the office. Before...

Continue Reading...

I just dropped back by work to take care of a few things and filled up a Coke bottle with water from the water cooler. I set it down by my computer and a couple minutes later went to take a sip only to find ice floating in it. It's not possible for it to come out of the cooler as ice, so the only explanation is that it came out so cold that it froze sitting in the bottle. Weird.

Apparently I didn't heed the warning at the bottom of the configuration page for the NTP pool that mentioned a bug with RHEL's ntpd. My score suffered greatly when ntpd disappeared around midnight.

The SRPM linked in the bugzilla post wouldn't compile cleanly for me, so I'm now running a self-compiled ntpd. Hopefully that'll stay up and happy.

This evening I set out to write a blog post on how different online mapping applications draw highway ramps. It turned into something longer than a normal post, so I've put it on my regular site: Looking at Highway Ramps.

Like I mention in the piece, I think ramps are a really important thing to look at in a map because they're something that's really hard to do right. I know that the code I've been working on recently doesn't solve any of the problems I list, mainly because of the data I'm working with. Always remember: your output will never be any better than your data.

Mostly unrelated: I think this is the first time I've put content on ericrichardson.com in about a year. I would say the web interface of the blog has drawn me in, but the frequency (or lack thereof) of posting here might dispute that.

Reading through some local blogs this morning I noticed ask's post from a few weeks ago talking about the ntp pool. That got me inspired to set up ntp on the new Cartifact server. I knew of a couple good local time hosts (particularly chronometer.usc.edu), and going through the list of stratum two servers I found one in the very same colo building. Judging by the whois on the IP space, I think that box is in CalPop's space, which is right across the hall from our box with Cyberverse. If I'm reading the stats right, though, that box is actually the peer with the most delay.

Our server, ortelius.cartifactmaps.com, is now registered with the NTP Pool, though it'll take a little bit before the pool decides it's reliable enough to give it out to people.

Since I got my new job, I've spent a lot of time here in the office. Some of it may be the work, but I think just as big is this view:

I love cityscapes.

The replacement drive for my laptop came in today. I installed Windows from my restore cd, used a Knoppix cd to run qtparted and free up some space and then went to install Linux via a Debian net-install cd I had lying handy (current as of a month or two ago).

The Debian install got through the cd part, but then things turned a little ugly when it came to installing packages. locale and base-config (I think... something like that) conflicted, and it wouldn't move forward. I flipped over to a terminal and tried to resolve the conflict manually, but only succeeded in making the setup process start respawning.

Handily, by this stage there's a pretty full-featured base Debian installed. I probably could have hand massaged package installation, but instead I used apt-get to install cdrecord, grabbed a new netinstall via wget and burned myself a new cd. That one's now happily installing.

There was a day when I used to think that sort of thing was fun. I think that day may have passed.

I went for a long time before I had my first hard drive die on me. Now, though, it happens with far too much regularity.

I got a new drive for my laptop in November of 2004. Now, just 16 months later, I'm watching Seagate's pretty cool SeaTools spit bad sectors at me. Seagate has a very handy 5-year warranty, so I'm covered there, but it's still just a pain to have to go through the replacement process.

Sitting on my desk I have a dead 60gig IBM Deskstar and a dead Maxtor that I can't find a size on (I think it might have been Kathy's). Both were manufactured in 2001. On the other hand I've got a drive sitting in the server that's from 1997. Heck, I've got a Seagate sitting right here that's old enough their system chokes on the serial number. Specs I found online are dated 1990. That drive still works. Why can't modern drives make it two years?

So I got back from the first bit of BarCamp LA four hours ago. Mack Reed has better coverage over at LA Voice, so I'll just leave my first day report at saying I enjoyed it.

But earlier today I was mentioning that I had a demo I wanted to get ready. Definitely wasn't ready to go at 4pm, but that's no big deal since I'm not presenting until 4pm tomorrow (err... today). And I'm proud to say it's ready. And if I do say so myself, it's cool.

I wish I had time between now and then to put together a cool presentation to go along with the demo. I may just hack something together in the time right beforehand tomorrow. Now, though, I need sleep.

So the LA edition of BarCamp starts in a few hours. I remember thinking it sounded cool when Sean made the announcement, but then I sort of forgot about the date. Sunday or Monday I realized it was this weekend and decided that I had a certain project I wanted to get into shape to talk about. Over the next few evenings I proceeded to get a little done here and there.

But now we're five hours before the opening and I've got 565 lines of Javascript and no backend.

I get the feeling this may be more of a Sunday afternoon demo than a Saturday one.

Definitely more here real soon. This project is a concept I've been sitting on for a long time, and I'm really excited about it.

This may make no sense to you, but just know that it's cool:

houston=# select AsText(the_geom) from buildings where id = 10;

MULTIPOLYGON(((
    3118897.96530012 13838758.1691906,
    3118951.74016637 13838841.8891787,
    3118993.06800402 13838848.7768498,
    3119057.74463768 13838808.25588,
    3119076.59247451 13838796.0172187,
    3119025.06695632 13838713.7759229,
    3118980.06661619 13838703.7761855,
    3118897.96530012 13838758.1691906
)))

houston=# select AsText(transform(the_geom,4326)) 
    from buildings where id = 10;

MULTIPOLYGON(((
    -95.3738909490995 29.7518641865238,
    -95.3737133770184 29.7520897224611,
    -95.3735825199263 29.7521051310476,
    -95.3733827457042 29.7519882370035,
    -95.3733245703251 29.7519529896557,
    -95.3734949130464 29.7517313269132,
    -95.3736376422065 29.7517076774085,
    -95.3738909490995 29.7518641865238
)))

Top data is in the NAD-83 Texas South Central coordinate system. Bottom data is your standard WGS-84 lon/lat. Using PostGIS.

I mentioned a few weeks back that I had made some JS code that crashed Firefox on Linux. If you're on Linux you can actually see this live now. Just go to the new Downtown L.A. Interactive Map, click on "Select a Search Type" and select "Residential Buildings". That'll pop up an interface to refine a residential search. When you select something under Type it loads some more options in under Price. Firefox 1.5 on Linux instead just bombs.

On a different site I'm working on (this one's a personal project, not a work thing) I've apparently made some javascript that creates a memory leak under that same Firefox on Linux.

Continue Reading...