The Goldfish Chronicles

A week or so ago, I noticed that a machine I admin had been rebooted 50 days ago, and another had been rebooted 10 days ago. “Bit odd!”, thinks I, “Who did that?. I’m enormously busy at the moment though, so I just had a look every now and then through the week to see if I could understand why. The two boxes that had been rebooted were Debian Sarges, while an OpenBSD next to them happiy reported 500+ days of uptime.

Tonight I had a closer look. I noticed some curious bits. Continuum thought it had been up for 45 days on a machine that had been up 10 days. In fact, top thought that it had had 12 days of CPU time, on a machine that had been up 10 days. Damn that’s impressive. The ps command was no help though, its idea of when things had started was pathetic, it thought Continuum had been up since 2007.

Same kind of ps issues on the other machine. Cron had been up since 2007, along with some others. Bizarre. Something messed up in the process stuff?

Then I had my brainwave. Each machine can be randomly screwed up, but it’s very unlikely that they can be the screwed up in sync. /var/log/dmesg shows the boot message for a machine - I’d thought it curious that each machine should still show the boot messages from 500+ days ago. So I compared them. The difference between their start times 2 years ago? 40 days. The difference between the surprising events? 40 days. Huh?

A friend of mine who’d been hearing all this in an IM (Thanks Jon!) did the right thing and had a quick google. Lo and behold, 497 days uptime is the day when your Linux OS will start lying to you - http://www.uwsg.iu.edu/hypermail/linux/kernel/0202.2/0337.html.

You live and learn. And curse and blaspheme and feel down for a week. As for the jiffies? I’ve no idea - some kernel thing and at 3am I’ve lost the mental focus to understand.

[Repeat 100 times: IANAL]

Binary and Source are interesting terms. When reading licenses I’ve always thought of Binary as the important word - something in a format that I cannot read. So I thought the following would be binary: .psd Photoshop files, .doc Word files. Even things which are technically text-ish seemed like binary: .pdf Acrobat files, .ps PostScript files. So naturally I’ve thought of Source as the opposite, things I can read.

That doesn’t fit if you look at things backwards. Source doesn’t mean readable, it means input. It means I have the ability to modify and (presumably) recreate. So .psd Photoshop files are source, not binary - the images are the binary. A .doc Word file is source (and has no binary, unless you count what you see on the screen or paper). A .pdf however is a Binary, if produced from an InDesign .indd format; but not binary if produced from something which edits .pdf’s directly. Using Fireworks on a .png becomes weird; suddenly it’s like an interpreted language because the tool uses the same format for source and binary - though the large Fireworks specific bits inside the png often are removed when producing the output, so that would then be binary.

It’s all very confusing. It seems as though it’s not just the source and binary files that count; it’s the tool. Imagine a code generator; its binary might be Java source code.

Antony Riley just released JarDiff, a tool for creating diffs between jar files, over at OSJava.

The history is that I was complaining about the slowness of the javap/shell script I was using for multidoc-jnr on our IRC channel (irc.codehaus.org#osjava; thanks Bob!) and Antony started asking questions. I had given Clirr a shot, but the need for dependencies to be available was causing me trouble (due to its dependence on BCEL), so he went ahead and created JarDiff, a tool that sits on top of ASM.

Here’s an example diff page from within multidoc-jnr: gj-core 3.1->3.0 diff. I like it, and it’s got the speed I need to throw it at the Jakarta Commons stuff again.

Odd part about moving; it’s time to dig the in-case-of-tornado emergency box out of the basement and eat the food. Beans, ramen noodles, tuna and ginger snaps.

We’re starting to pack for the move to Seattle. I love being a bookworm, but they take up a lot of room when you pack them. I’ve used up the 5 boxes I had sitting around and I’m only halfway through the computer books. SciFi books, Carrie’s books, various others - probably another 20 or so boxes minimum. 30 boxes of books. Damn - I really, really need digital paper.

I’ve been reading Peter Morville’s Ambient Findability recently. Peter was one of the two authors who brought us the seminal Information Architecture from O’Reilly, even after all these years - my favourite IT book. Incidentally Barnes & Noble have it available for very little money in their sale section. Buy, buy, buy. One of the interesting and not fully realised parts of that book was the concept of search and navigation; mostly because when it comes down to “where are you going to put stuff?”, “how are you going to find stuff?” is an enormous part of getting to the answer.

One thing, among many, that Peter talks about is RFID. I’ve not finished that chapter, let alone the book, so maybe Peter will point this out, but with IPv6 meaning that every object on the planet can have an IP address, and RFID/GPS and other technologies meaning that every object on the planet can have a lat/long, I’m starting to get scifi deja vu. Let’s pretend that every RFID can be looked up in a database for its avatar; or maybe RFID-2 allows larger objects to carry their own avatar (for all I know RFID allows this already). This means that one can use a browser (or VR environment) to experience a matrix-like cyberview of anywhere on the planet.

I’m sure there are lots of smart people out there talking about this, and iterating the actual steps needed to get there, but this was my “hey it could happen” moment.