The Grand Misadventures of a typical wavefunction v.v

champaigne

2008-02-27T16:14:34Z

It's always painful going back to your code when you've learnt a lot in the meantime. Or maybe it's just real coding that's painful.

I seem to not be good at anything these days. I'm simply OK at a lot of things. I get the feeling that might be changing, though. I started learning twisted yesterday, and that's something that seems to be quite employable. Getting a job doing something like that would no doubt lift my hopes. And maybe that was that spark that lead me to play piano tonight- so much fun. I've played it a few times since I've been here, but each time it's felt like a chore. Last year was hard. Maybe not as hard as the year before it, but it was still hard. And I really need to break out of this. I need to focus.

Actually, I was going to say something informative and interesting, related to le CAPTCHA, but by the end of it, it didn't seem that serious, and the punch line was really missing. I guess I'll just get back to hacking, and let you know when there's something I have for interested parties to play with.

Love.

Hurd!

2008-01-09T14:30:00Z

I thought I should write down the steps that I recall I went through to get a working Hurd installation- I may need to do it again someday :) and it wasn't exactly trivial, especially to a n00b like me.

I'd tried both the k14 image and the l1x image. Unfortunately, the l1x image did not work for me. I never spent the time to work out why [the init script calls fsysopts to set the root filesystem writable, but it would never return, even if the filesystem was already writable. I gave it up and went back to k14]. If I were to start over, I'd probably use Xen, for reasons I'll address later. If that's the case, you need a PAE disabled hypervisor. There's also the qemu option that many people are using these days, and yet others still are using vmware. Some people cross compile, though I hear it is not for noobs like I.

Before you do anything, you're going to want to make sure you can run the Hurd on your machine. There's a hardware list somewhere. At the time of writing, to run it natively, you're going to want an older x86, without SATA (or at least with an ATA emulation option in the BIOS), less than 3 GB of ram, and no USB peripherals you couldn't live without. It's also pretty pointless using a shiny new graphics card, since they all require kernel modules. If you are going to be using a virtual machine or emulator for the Hurd, you're not so limited, but your VM will have to be.

Also, the Hurd currently takes no advantage of SMP or 64 bit processors, though there is a port to the latter underway. Put simply, you will feel a bit silly running it native on any recent machine :)

Installation from kx images [k16 being the newest at the time of writing] is pretty straight forward. The CD boots into a menu that allows you to modify your partitions and install the kernel and base distribution. It doesn't set up grub, however, which is required. I used some of the features on the lx1 CD for setting up grub, although I installed it natively, following the instructions on the GRUB wiki. Once you've booted into GNU in single-user mode, you will need to run native-install, reboot and repeat. After this, you will be able to reboot into GNU in multi-user mode.

The installation instructions mention the console [more correctly, the console server], but they do not mention that it is required to run X [as I understand it, the repeaters it starts are required to run X]. I used the gui.sh script on the lx1 CD to set up X, but I don't think it did very much- I'll need to take a look at it. I still had to get packages containing the driver for my graphics chip, and probably input devices, and various X server packages, in order to run it. Then I used dpkg-reconfigure to set it up, and installed fluxbox.

The base distribution is pretty slim [I heard 160 MB?] so you need extra packages if you want to do anything interesting. I should get a list of what packages I have installed, I guess. I remember getting emacs, xchat [which automatically got me gtk], rxvt, wget, lynx, build-essential, gnumach-wireless-tools, and that's really all I remember. I'm attempting to install pidgin soon [from source], and I attempted to install Firefox 3.0b2, but Michael has told me that XULrunner hasn't been ported :(

It's a shame that Firefox doesn't work. Lynx isn't really a substitute. I couldn't find a pdf reader that had a package either, so there are three [along with music editing, of course] things that I need to keep GNU/Linux around for. For this reason, I'm thinking when I get my new hard disk, I'm going to use Xen. That way, I'd be able to have the Hurd there for hacking, and have GNU/Linux there for general browsing and music work.

Fleeb

2007-10-25T15:03:51Z

Ugh. I am really, really sick of this course. Just in time, I guess.

Deep in exam-preparation mode, but not in the zone, because it is just so boring! Linear models and regression tests and t-distributions and F-tests.. ugh! Statistics statistics statistics! Where is the Math? Time to go kung-fu on its useless ass.

Someone save me!

ps.. seems Firefox does not have the Emacs nature. Hmm.

does firefox have the emacs nature?

2007-10-23T04:35:44Z

touch ~/.gtkrc.mine
echo "gtk-key-theme-name = \"Emacs\"" > .gtkrc.mine

Mmm

2007-10-20T18:53:42Z

Well, that is it. With some help from a fantastic friend, and a few hours of work, I've got my major project done and in [at least, email copy, I've still got to drop a hard copy in an inbox tomorrow morning]. I can only wonder what will happen next. Will I need to apply for some kind of special consideration in order to get good marks? All I know is, I feel a lot more relaxed knowing there is little more I can do. Anyway, one, much smaller project to get in tomorrow, some beers later on, and then it is time to prepare for exams. Yay. At least I can't say that nothing is happening, that things aren't going anywhere.

I need hugs, and beer, and sleep, and things.

On the horrible suckyness of paywalls when you're actually trying to do research

2007-10-20T09:44:26Z

William Leslie, et al.

We present a method to absolutely confound any individual trying to do research. We consider the effect of not letting people view the results of a three year drug binge that is the "PhD", and watch the impact of this on people trying to do get timely information for the assignment due.. today.

While paywalls have been studied before [Stallman, Lessig, et al], they have rarely been visited with me in mind. We set me up, requiring peer-reviewed content, in order to finish an assignment. Moreover, we have made me sleep an extra day, so that I do not realise that today is in fact the day I've got to get this in. Then, we watch as I struggle and scream, trying to get that last bit of useless data for my assignment. We see that the pay-wall is effective in stopping the spread of information to people who really don't care about research, but need some, damn fast.

Finally, we present the results of me handing in an assignment late. Again. And hoping it is accepted.

Click HERE! for full article and pay me $10.

stress stress stress.

ekey, Rhianne, Rant, + Hurd update

2007-10-07T17:02:52Z

Yay! I have the Hurd running peachy. Installation of the default X environment failed, though, so I need to hook that machine up to the 'net and apt-get me some xdm. Haven't bothered putting FreeBSD back on that machine yet, but it is in the works. By the way, if any of you are thinking of running the Hurd, don't bother with LX1, it uses fdisk, which is horrible. It is supposed to move to cdisk, but there aren't any images floating around. Just use K14 and hit me up on msn or irc [usually on freenode/#Hurd these days] for help.

Oh, and I made my first posts on the Coyotos-dev mailing list. Getting to chat with the brightest minds in kernel design on the future of that project really made my day. I'm not sure I'll ever be a productive Coyotos hacker -I'm certainly not up to scratch on it yet- but damn, I'm having fun.

My cousin Rhianne got married yesterday. Weddings are so sweet. It was good seeing lots of my family again- family I hadn't seen in a few years at least, and five-ish years at most. Rhianne and Eddie have been together for about eight years, and I can't even imagine what that must be like. Speaking of me, it was a bit of a cause for reflection. I have been so long without love [in fact, I'd be willing to call it two years] that I can't really remember what it was like. It doesn't help that I'm so walled up these days. I've noticed that I seem to [it seems almost intentional] destroy relationships that I have with friendly, attractive females that actually get close to me. I've known this for a while, I guess. I don't really know what to do. Or even if I really want to do anything.

Along the same lines, the guys in #python were talking about their wives the other day. The impression I got was to just keep doing the things you love, because you're bound to bump into girls who dig the same things you do. It's not like I'm terribly picky that way, though. There are some things that I really require before thinking about dating someone, such as empathy. The person who has taught me most about this is Jane. She's great example of how never to get into any meaningful relationships. She will never tell you when something is wrong, and when asked, will never tell you what is wrong. She will just be stroppy for days. And the terrible thing about that is, she's like that all the time anyway, because of school, her business, feeling sick, having to do the dishes while I'm away for a few days etc, so there is NO way to tell if she has a problem with you. It'd be nice if girls that play stupid games could be marked with a big "not worth your effort" sticker.

The funny thing is, none of my close friends are like that [depending on your definition of close, maybe one], but you always hear about people like that. I've been very lucky not to date people who play stupid emotional games. The closest I got couldn't really be blamed because she was fourteen at the time and it really was my fault.

Anyway, back to selectivity: once I've checked person.gender = girl and person.isasshole = false, I'm pretty open to being wowed by anything. I've met a lot of really amazing people, but I seem to meet them pretty regularly as well, which suggests that even if the bar is high, there is a steady supply. She doesn't have to be a Mathematician or other scientist, she doesn't have to be a musician, she probably doesn't even have to be an artist [but that's a hard one to avoid if you're cool]. In fact, even if you were a stunning redhead who rides around in a teacup* and had no other interesting features, [though it is hard to imagine,] I'd be smitten.

Of course, that is compounded by the fact that I am not interesting. To anyone but kernel hackers. Now there is a huge pool of single girls for you. At least they are geeky.

On a lighter note.. I really wish I had a key that said "Emacs, Dammit!", that I could bind in fvwm to map the keyboard on to the proper Emacs commands. Ctrl-v should be page-down EVERYWHERE. At least, with an "Emacs, Dammit!" key I could hold ED-C-v to scroll down, without taking either hands far from the home row. What would be better of course, would be all applications using Emacs keys were appropriate, or configuration items to have text input automatically treated with Emacs [or Vi, if you're evil] key bindings. But right now, I'd be satisfied with an E,D! key. [practical detail- I could probably bind something like the left arrow key to do this. It is reachable on my keyboard with my right pinky, which leaves most typing unhindered. Ideally additional chording keys should be below the space bar, because thumbs are stronger than pinkies.]

Oooh, and I've started writing jazz music and looking at tango music. :)

Ps. I'm so far behind on my Lorenz Model project. Someone make me a time machine.

*if you know anyone that fits this description, point them in my direction please. Thank you.

With a fighting chance!

2007-09-27T12:03:07Z

"as far as I know, that the older Thinkpad BIOSes took partitions of the type n * 16 + 5 (for any n) as DOS hibernation partitions; FreeBSD slices have the type 165..."

Lol, tricky. So:

Get a floppy drive working [ancient technology lol]
Create BIOS update disk
dig up Thinkpad floppy drive
remove Thinkpad hard disk [hibernation slices override boot order]
Boot the floppy and flash the BIOS
Replace HD
get on with fixing the GNU!

Link 1, 2, 3, Custom kernel parameters

Thanks to KORBEN and the FreeBSD team for tracking this down. I can't imagine how hard it must have been the first time around XD

Freedom 0, tiredness 1

2007-09-26T15:00:21Z

Installed the Hurd this afternoon.

Broke it. I'd pulled the plug, so my root filesystem didn't get umounted properly. Now I need to boot to something else so I can fsck it. But, Knoppix is not working. FreeBSD is not working. In fact, even my BIOS seems to be acting strange. Considering trying Damn Small in the morning.

No-one said freedom would be easy. Especially when it's still experimental.

+ooh, captcha will be using cairo, now. This should make some things a bit easier. Cairo also has neat python bindings, if I didn't have a fair bit of C written I'd write it in python now.

urls go wlid!

2007-08-17T14:36:29Z

Please be careful with Windows Live ID. Same with OpenID, I guess.

Anyway, just wanted to point out how cool Coyotos is. Found an interesting post on the mailing list from way back..

..should more or less gracefully survive the following
disruptive situations when home and OS are on separate disks:

Crash of disk with operating system and complete reinstall of OS on the new disk possibly with other geometry.
Crash of disk with home directory and recovery of home directory from tape.
Reinstall of OS because of change in processor architecture (For example, x86-32 => x86-64) when new OS supports programs for older processor architectures (for user’s process).
Replacement of mail server on other one but with compatible mail handler interfaces.
Replacement of mail hander program.

I don't even usually think about this stuff because it's not what I'm aiming for, it's not what I care about. But it'd be pretty damn awesome to have the power go out and your computer start back up instantly just the way it was when it went down. It's the weird stuff like this that makes it seem like something cool to hack. Come on, an operating system that stays up while you're replacing the processors?

Started actually coding the captcha today, only got about a hundred lines done. It IS a start.

who | grep -i blond | date cd ~; unzip; touch; strip; finger; mount; gasp; yes; uptime; umount; sleep

Need Distractions

2007-08-09T14:41:23Z

Eh; things move so fast and yet so slow. Starting this week I'm out of work, which is not much fun at all, though admittedly my last job wasn't as much work as I'd like. I'm not sure if this is the best time or the worst time for that contract to end, as I was hoping to start looking for somewhere new to live this week.

Anyway, it appears there are plenty of jobs for programmers around. sure, most are Java or PHP, but there are plenty of well paying C/C++ jobs around. It's going to be tough, but I guess I'd better get some experience in that area over the next week, and it's a lot easier than getting another engineering job [because I'm not actually qualified to do that :P]. So I think the next week had better be spent coding.

One of the things I've got to do is work on the CAPTCHA for the Nessness CMS. Lately I've been working on a sort of library it uses for drawing smooth curves between points, which will most likely be a separate thing, as I can see it having uses outside CAPTCHA. It's a way to imitate handwritten text- common variations in writing style will be supported, and the text and noise in the CAPTCHA will look just like neat or messy handwritten text, with the hope that such variations will be difficult for a computer to decipher [though that is only the beginning of this CAPTCHA.. :) ].

With the move to coding I can finally move my laptop to GNU :) Though the idea has had me on a high note for a while now, it is looking like it will be a bit of torture. Supposedly only old window managers work- not even my lovely fluxbox has been ported. So it looks like I'll be keeping a FreeBSD partition around for a while [as well as a MINIX3 install, too, I guess, just to tinker with]. Something about the way threading is done under the Hurd. It will be interesting to see what works and what doesn't [like wireless support, GASP]. And even more serious: WINE. Because it's what I'd like to hack first, once I get the old projects off my back, and if I can stay away from the kernels.

The other thing that is bugging me is my car. Working from home is one thing, but not having a car when you need a job to move out is a huge pain. If I can't figure out what I need to do with that diff this weekend, I'm going to pick up another Saab [there's a really nice hatch on Egay going for just over a grand]. If I do go that way, though, I'll miss out on being able to say that I bought a really snazzy car for $400 + rego + some tinkering, which is the main thing [sort of like it is with the GNU :D].

Aside from that I am not well which is pretty upsetting. I couldn't make it into school today, and we had an assessed class :(.

Speaking of school, I tried to do some homework tonight. You know, Maple is a funny thing. On one hand, it has some really powerful symbolic manipulation, like working with infinite series and continuous functions. On the other hand it lacks basic things like objects. It also lacks basic tensor manipulations [on things other than tensors, which is where you need them anyway]. I spent about half the day trying to get a sum over a vector to work. Indexing a vector works fine until you try to sum over the indices. I checked and double checked, it's just some Maple stupidity. [The answer is to dot it with a vector full of ones. Yay. This is a lot neater looking than a map, sum or seq, but it's kind of silly.]

I realised today that the function seq() is really quite general. If it could work like the lispy map(), and in doing so could handle tensor-like index manipulations, it would be unstoppable. One of these days, -one of these days that I'm not really busy, ie, never- I should reason about what sort of functions the IDEAL language would have. The old Python from __future__ import knowledge would be on the top of that list.

Heh. Anyway, so tomorrow I've got to get that assignment done, which isn't much of a task, BUT, I don't actually remember any of the maths, so I'm going to have to kind of make it up. I should do kind of OK anyway, it just might take me all day. But that's ok, because maths is fun. And other than that, I still haven't handed work the final drawings I've had to do. So that's tomorrow covered, I think.

If I go to bed now, I might actually get up at a reasonable hour.

Absolutely

2007-06-16T07:18:56Z

hateful_monkey (1081671) of slashdot had this to say. The context was about scientific consensus and its effect on the global warming debate. I think it is a very good point that he makes:

"I think the point that many people are trying to make is that suppressing the minority view by shouting them down or dismissing them is crackpots is rarely a productive stance. No one cares if someone claims global warming can be stopped by wearing underwear on your head unless enough people begin to believe it. Some are beginning to make noise about stripping dissenting scientist of certification and awards if they publically denounce the concept of global warming. Why? Because each "anti-warming" voice gives a little extra support to people who would rather not address the issue because of greed, apathy, our political expediency. The dissenting voices may truly believe that global warming is not happening, or is not caused by humans, and in some cases they may have data to support their case, but their reasonable and well thought out ideas are used in broad and irresponsible ways to serve political ends.

The debate is not being driven by opposing scientific opinion, it is being guided by self interest on both sides. Scientific consensus, like all consensus, is sometimes wrong and should not be allowed to protect itself from criticism by shutting down the debate. The same is just as true for evolution, general relativity, and quantum mechanics, as it was for the Earth-centric model of the Universe, Alchemy, abiogenesis and the hundreds of other "scientific" ideas that have fallen by the way side. Someone will always point out that many failed "scientific" models failed before rigorous use of the scientific method was common, and that is true, but the scientific method REQUIRES dissenting ideas to test against or it is nearly worthless."

[formatting mine]. I quote the post in its entirety, I do hope he doesn't mind.

Scribble Scribble Scribble

2007-06-06T04:54:44Z

Here's something kind of cool. It makes me think though. Mostly about how wasteful programmers can be. My first write of the graphics end of my port of flOw manipulated bitmaps in a similar fashion, though it also required rotations, and while it was hand written ASM, it was running well over 20 fps. I didn't notice at the time, but had I used MMX registers, I could have quadrupled that [as long as I used 24 bit colour or less]. I opted for a better option using OpenGL and vector graphics, because it would allow me to make the game 3-d later too, but the point still stands: we should see computers doing this sort of thing. They certainly have the power, just not people who can write algorithms.

I'm not suggesting anyone write assembler, but I notice, say, that context sensitive menus in Firefox take half a second to open. That's really annoying. All it's doing is displaying a box with some text in it, where the text calls a function. Menus have been around since Xerox PARC, how can they take half a second on a machine today? It's not like there are a lot of contexts to consider, or anything. I can't work out where all that effort is going.

Right click a heap of links in succession in Firefox to feel the slowness!

It's worst when a tab is still loading while you right click. Firefox can't think about your menu until the tab creation function is finished. Why? Bad threading, I guess.

Actually, I could gripe a lot about Firefox. While I understand that Gecko is complicated because, for example, M$ frontpage creates broken HTML on purpose [and even expensive apps like ColdFusion make a mess of Javashit], surely at some point in the design someone would have noticed that different tabs, or at least different domains, need to have different javascript namespaces, most likely different threads, and most probably different processes. But this context menu thing is just crazy.

And while we are on Firefox, has anyone got any idea why there's no context menu in the history, bookmarks or previous to allow you to open those pages in new tabs? I'm sure I could write a book about it, but it'd be much faster to simply fix it, so I digress.

There are several code bases that I'm half working on that all have really terrible programming. I think there's this common theme these days where people are using OO, but not for the right reasons. While multiple inheritance is great for code reuse, the real issue that OO was supposed to solve was that people tend to think too much in terms of functions rather than data. When writing something, it's normally best to think about what sort of data is required to completely describe the state of the system. Then, your functions become somewhat obvious- moving from one state to another, or using the state to call I/O functions.

I think that this way of programming also leads naturally to some Parallelisation, for example, it becomes obvious in a function when you're operating on separate chunks of data.

---

Now that the last five minutes gripe is out of the way, I can scribble down some things I need to do. I've been looking at a memo out of MIT, called HAKMEM. It has a heap of nifty problems, I didn't get very far before I saw problems that I can solve. Here's a thought on simplifying large multi-input logic gates.

Binary under XOR and AND form a field. It's the field Z_2, or integers modulo 2 with the obvious multiplication and addition. Now 'OR' also distributes over AND, but OR is not invertible, so it cannot form any field-like structure. However, we can write A OR B as AB + A + B in the field above. I think the question for me here involves a minimum number of two-input logic gates, and I think we've almost answered it by thinking about this field. We only need consider factorisation of multivariate polynomials [which are at most first order in all variables], and we've almost got our minimal representation. Because XOR distributes over AND, we can pick out XORs in the factors that are most common, and take them outside the ANDs. Statistically, once we've got half of the complexity in XORs and half in ANDs, we've got a well rounded tree structure, which presumably holds the simplest representation.

I've also got a reasonable amount of data and functions drawn up for flOw. I think the thing that is staring me in the face here is writing the OpenGL part, because it's going to take tinkering. What if my baby is ugly? Granted, I can make it look however I like, but I'm kind of scared, like I'm painting with a blindfold on.

I've been reading the specs for a capability based microkernel, Coyotos, and of that poster child for hardware design genius, the DEC Alpha. As always, I've taken issue with this and that, and thought about building on those designs. Coyotos in particular has a number of invariants which I don't feel need to be explicitly stated, which would not be required had the design been completely holistic. That said, I still feel like working on the Coyotos project. As for Alpha, it really is something! But I'd like to pick at some of the cache coherency stuff and OoOE [not that I've seen the real OoOE spec used on the later Alphae], just things I think could be done better.

Albert, being the cleverness that it is, has taken this in too. We've now got better pipelining in the works, including ways to divide up the processor for vector operations. I'm at a bit of a loss as far as division goes, though, and might just have to leave division taking O(lg(n)) cycles. This isn't such a bad thing, I guess, but it compounds data dependency. For example, there are two ways to calculate the inverse of a product of m variables, and one way takes about m + 567 instructions, and another more obvious way takes 512m. as you can see, for m = 3 there's a significant difference. Have a look what happens for larger m.

I guess those sort of problems can be overcome reasonably using an optimising compiler. Considering the sort of parallelism we will be looking at, it's probably not a bad idea to assume someone is using one. A large part of keeping all the machines busy will be OoOE and dynamic dataflow analysis, and allowing instructions that call for help with obviously parallel algorithms, such as map and reduce. We've also got branch hints, and the concept of there being one correct data form- that dirty bits are broadcast [I want to broadcast your dirty bits! LOL] so any op that relies on that data can find the correct version. But optimising is still important, because a machine will only be able to look so far ahead for data dependencies without a performance hit.

Minimal WINE for Newbies

2007-05-24T19:05:28Z

So, that geek chic friend of yours [you know the one, who accidentally wipes your hard drive and then gives you no help actually fixing your computer] has installed this new Linux thing on your computer, assuring you that it's the way of the future and it's more ethical than what you were using before. You've poked around and worked out this permission stuff and you know where to go to get bash help. Now you want to run your Windows applications.

Almost all Windows applications CAN be run in Linux, with a program called WINE. For someone not used to Linux, WINE can be a little daunting. There is a tutorial on the WINE homepage, but it's really quite lengthy, so I'm going to try to give a more simple description on how to use it that will work with most Windows programs.

What is WINE?

WINE is a recursive acronym for "WINE Is Not an Emulator". WINE translates Win32, Win16 and DOS [and maybe soon 64 bit Windows API] system calls into POSIX and X ones [used by Linux and X]. It also translates DirectX instructions into OpenGL ones. So it pretty much translates the questions that Windows programs ask into questions Linux understands. Because WINE is not an emulator, it runs programs with about the same performance as they would under Windows [sometimes better, sometimes worse, depending on what exactly the program is doing]. WINE also contains no Microsoft code, so you don't need to be bound by some stupid fifty page EULA.

Installation

If you don't yet have WINE installed, it's usually quite simple. You can check your package manager and go from there. However, some newer games and such require the newer WINE features, and the WINE packages in some distributions get dated quickly. If you think you need the latest version, you may have to build it. If you've never built from source before, you're going to need a compiler and the default headers. Open up your package manager and find gcc and gcc-dev and install them. You may need to make sure you've got the latest libc6, again these should be in your package manager.

Go here to get the source for the latest WINE. Download and extract the tar.gz file into your /tmp directory. Open up a command line [as root] and go there (cd /tmp/wine and press the tab key and then enter will give you the correct directory in bash). Type "sh ./configure" to configure a new version of WINE for your system. If that works fine, type "make". This will take a while [forty minutes by my count], so you might want to plan around that :) when that's done, "make install" should finish the job.

Now you've almost got a fresh WINE installation on your system. Open a command line as yourself now and type "winecfg". It will set up a fake Windows account for you. You probably won't need to mess much with the settings here yet, but have a poke around anyway.

Installing Windows programs

Once you've got WINE, installing Windows programs shouldn't be much effort. WINE supports most installer types, and so you can just run them by clicking the icon. On a CD, this is usually something like "setup.exe". If you want to run your installer from the command line, go to the location of the installer and type "wine setup.exe" or whatever the name of the executable file is. Install should proceed as expected in Windows. Make sure you install in a location you have write access to. The fake C: drive WINE sets up for you will work.

Once the install is finished, you may or may not have a working Windows program installed on your computer. You can find out by trying to run it :P. If that doesn't work, don't worry! There may be a few other things we need to do. Try copying all DLLs the program has installed into the program's main directory [try checking C:\Program Files\Common Files to see if there are any DLLs there]. If WINE is still complaining about DLLs, check that you've got all the DLLs it's asking for, if not, try searching Google or your old C:\windows\system32 for the DLLs. If you're still having problems, try messing about with the settings in winecfg. Some programs run better when WINE is pretending to be Windows 98 for example, though a reinstall of the program may be required. Some programs run better with certain mouse or 3d effects toggled. Explore the settings in winecfg.

By the way, the fake C drive refered to in WINE is usually in the directory ~/.wine/drive_c. If you can't find .wine in your home directory, make sure you can see hidden files [it's usually a menu option in your file manager].

This should be enough to get most Windows programs working under WINE. I know it sounds like a lot, but chances are it won't take more than a minute to get a program to work, which is a small price for a clear conscience and piece of mind.

The Perfect Girlfriend

2007-05-09T05:22:03Z

Stolen from Robbeh!

Physical Characteristics

Preferably at least one of Glasses, Braces and straight clean teeth.
Healthy to moderately lean, please.
Fine features.
Soft, clear skin. Not distinctly tanned. Soft also implies body hair should be removed regularly.
Terrible haircuts are a no.
Vagina.

Personality

Empathetic.
Stable. Drama queens need not apply unless sufficiently hawt.
Straight would be nice! Not required, though.
Intellectual, interest in either liberal arts or science is a plus.
Not overly political, but well studied and considered whenever it comes up.
Picky, with who she falls for. Won't fall in love on the train every second Tuesday.
Eccentric and social, but not too social not to turn down being social for more time having sex.

Career Prospects

Doesn't stay in a dead-end job without good reason. A sufficient reason would be because she enjoys it and it doesn't ask too much of her time.
Artistic, research or executive job prospects are best. Large, clean mahogany desks in corner offices are considered a big plus [for sexual activities].
Her career should not be her life! Even if she's doing something she loves for a living, if she suddenly decides she'd like doing something else, she should have the guts to do it.

Family Prospects

Would like children sometime. That's not essential, though. BTW, "sometime" is sometime several years from now :P
Knows how to treat children [with "empathy" covered, I think we're mostly ok].
No white picket fences please. Having your life planned out for you is not much fun.

Until next time, here's something you can do with a dead badger.

:)

2007-05-02T08:29:56Z

I've come across something probably relatively minor but it feels important to me, so I felt I'd blog it in my tech blog, which I will do later. But first, I wanted to start the story here. In the first semester of second year, we took a subject on computer programming for mathematicians. It was divided up into halves, one for C, and one for MATLAB. Despite my reservations concerning programming, I quite like the idea of C, there's some sort of raw power in it. The machine does exactly what you ask and nothing more. MATLAB, on the other hand, is a poor mix of perl, C, java and maple. It is supposed to be powerful, but I can't find a use for it. Besides, it is made for *gasp* engineers. Yuck.

One of our CRAPLAB tutorials involved using a sieve of Eratosthenes. It's an ancient algorithm for building a table of primes. It's very simple, you start out with a table containing all the numbers starting at two [well, up to a certain number], find the first number that isn't crossed out, and cross out all multiples of it. The numbers that remain when you've reached the square root of the total numbers in the table are prime. At least, that's how it should work. The implementation instead removed numbers from the table if a built in function, "isprime", failed.

I'd never heard of the sieve of Erastosthenes, but I declared the algorithm to be broken anyway, under my breath. I needed the marks so I did the tutorial and left. Later in the day, I came back to the lab to find Sam and Bambi putting the "sieve" through its paces.

They were trying to outdo each other in calculating the larges prime table. In the tutorial, we were explicitly told not to try any more than fifty numbers. I believe Bambi had just decided to go for broke and try to calculate all primes less than a thousand when I showed up. Of course, being the concerned but egotistical math student I am, I pointed out the flaw in the algorithm. I explained the sieve, how I thought it should work, in greater detail. My waffle was received as "interesting, but I don't see it working". You know, kind of like people with Linux.

I quickly logged in and fired up maple. Sam was calculating the primes less than two thousand, and Bambi started his quest for primes less than ten thousand just as I typed my first line of maple, which is, for historical reasons, "restart;". I used only binary values in my implementation. Before finishing my typing, Sam's calculation completed and he embarked on calculating all the primes less than twenty thousand. We all remarked that considering it took over seven minutes to calculate all those less than two thousand, this calculation may not finish.

I quickly tried my code. Missing semicolon, la-di-da. I tried calculating all primes less than ten thousand. It took less than a second. I tried calculating those less than a hundred thousand, and it still took less than a second. A million? A few seconds. In less than a minute, I guess, I'd caught something of a computational mathematics bug. Now I don't really consider computational mathematics to be that much of a mathematics. Sure, there is some very heavy maths there! But it's just not smooth and perfect. Nevertheless, it was something that would stick with me.

I pashed the antelope by a factor of ten again. This calculation took a few minutes. This didn't make any sense at first. Ten million is not a huge number for a computer. But it actually is in maple. What was worse, we were running windows, and it reached the point where it swapped like crazy. We have a Unix server available in the lab too, so I considered moving my code to C. I dropped by Dr Andrew Stacey to find out what sort of memory concessions I could make and eventually decided it wasn't worth my effort anyway. That my algorithm ran that much faster than Dr Hill's CRAPLAB algorithm made my day.

I never really stopped thinking about primes. I'd love thinking up new ways to extend the algorithm. Tonight I came up with a new extension that I like. I figured I'd only need to store values that are equal to one or five mod six. I calculated the new algorithm for crossing out the multiples and realised that it is probably best to push all complexity into this. I haven't calculated weather or not that's true yet, because I haven't really generalised the algorithm like that, but I do know how to generalise it. And that's what is neat. It relies on the Chinese remainder theorem, and it gets rid of one O(n^2) step in my original algorithm. Not only that, it expands the size of the working set from n to at least n.ln(n), allowing more primes to be found before flushing the memory.

Explanation and complete algorithm to come soon on the tech blog.

Interesting hypothesis:

\[A_0 = n,
A_i = A_{i-1}*\ln{A_{i-1}}
=>\lim_{i->\infty}A_i = n^2\]

?

On the origin of the species: Laugh 1

2007-04-21T21:55:09Z

"Changed habits produce an inherited effect as in the period of the flowering of plants when transported from one climate to another. With animals the increased use or disuse of parts has had a more marked influence; thus I find in the domestic duck that the bones of the wing weigh less and the bones of the leg more, in proportion to the whole skeleton, than do the same bones in the wild duck; and this change may be safely attributed to the domestic duck flying much less, and walking more, than its wild parents. The great and inherited development of the udders in cows and goats in countries where they are habitually milked, in comparison with these organs in other countries, is probably another instance of the effects of use. Not one of our domestic animals can be named which has not in some country drooping ears; and the view which has been suggested that the drooping is due to disuse of the muscles of the ear, from the animals being seldom much alarmed, seems probable."

-On the Origin of Species by Means of Natural Selection, or the Preservation of Favoured Races in the Struggle for Life, 6th edition: EFFECTS OF HABIT AND OF THE USE OR DISUSE OF PARTS; CORRELATED VARIATION; INHERITANCE.

In other words, the domestic duck grew stronger legs and weaker wings because it stood on the ground, rather than the domestic duck stood on the ground because it had weaker wings and stronger legs. Likewise, the cow grew large udders because it was milked regularly as opposed to people milked cows because they had large udders.

Exactly why this did not occur to Darwin I am not sure. I am reading this with a strange feeling, somewhere between reverence, concern and confusion. I've problems with evolution and I've problems with creation [not to mention ID is a crock] in the ways they are usually spoken about, and further to this, I think debate on the subject misses most marks and fail to take certain circumstances into account, but without a deep and concise understanding of the matters at hand I have no method by which to deliberate on the facts and considerations.

Using Python

2007-04-18T05:50:03Z

Maybe Eric S. Raymond's crazy writings on Unix have scarred my brain, but I think that every computer user, no matter how casual, should have some coding experience. Even if it's just shell script, for example, it's nice to be able to use basic tools like grep, wildcards and pipes. You might not use them that often, it's true, but the moment you get stuck with something difficult you want to do, it sure is nice having good tools.

And I guess that's what Python is to moi. I wouldn't use it for real programming [games, maths etc], but when I need something done and need it done now, Python is the tool for the job. It's also pretty good for that web sort of scripting, on the server side. It's better with strings than Java and it's better with numbers than C, so it's really quite powerful- just not that fast. That said, the following program runs through a CD in under half a second, so it's fast enough. Just don't go inverting any matrices.

Ok, here's the first program, makelist.py:

#For creating a list of files contained in a
#directory and its subdirectories.

import os
filelist=open("C:\\diskfiles.txt",'a')
allDirectories=os.walk("C:\\My Music\\")
for directory in allDirectories:
   filelist.write("\r\n".join(directory[2]) + "\r\n")
filelist.close()

Or, in plain English:

Load tools for interacting with the operating system,
Open a file to save the results to,
Get information about the files,
For each directory,
- get the filenames,
- join them into a newline-delimited string,
- write them to the file,
Close the file.

Once we've got some nicely sized lists of files to compare, we've got to compare them somehow. There's several thousand entries at least, so we will use a program to do it for us: comparelists.py.

#For comparing lists of files on backup media to
#files on disk.

saved = open("C:\\savedfiles.txt",'r')
onDisk = open("C:\\diskfiles.txt",'r')
savedL = []
diskL = []
for line in saved:
   savedL.append(line)
for line in onDisk:
   diskL.append(line)
savedS = set(savedL)
diskS = set(diskL)
notSaved = diskS - savedS
notSavedF = open("C:\\notsavedm.txt",'w')
notSavedF.writelines(notSaved)
saved.close()
onDisk.close()
notSavedF.close()

open the files we've made.
make some empty lists to hold the contents of the files.
read them in- we're doing it one line at a time in a "for loop"
make sets out of the lists. This gets rid of duplicate elements.
Select the files that are on the disk and not on any CDs.
Write the output to a new file.
close all the files.

heh

2007-04-16T21:14:50Z

So, let me explain how things are. Over the years, you've been running out of space on your computer, and you frequently copy things to CD- but you rarely keep track of what you copy and move about, so you're not too sure what you can safely delete -ie, what has been backed up-, and what you would need to save if you're going to remove. What you need is something to read all the filenames on the cds and all the filenames in your directory and compare them for you. You don't want to know about things you've already removed, just things about files you have not backed up yet.

So it's a problem you can state in a single sentence. Clearly, there are different ways to do this. The typical user might just give up. Especially one not used to having a shell, the power of a command line interface. Anyway, so here's my first attempt at using unixy-power under Windows [where most of my data is]. It's a skeleton written in Python that needs to be modified to suit my needs [perhaps interpreting the filenames from sys.argv?], but it's lovely none the less. And it's what, 22 lines? I have no idea how much effort something like this would be without python.

#For making the lists

import os
filelist=open("logfile",'a')
allDirectories=os.walk(pathToWalk)
for directory in allDirectories:
   filelist.writelines(directory[2])
filelist.close()

---

#For comparing the lists

saved = open("",'r')
onDisk = open("",'r')
savedL = []
diskL = []
for line in saved:
   savedL.append(line)
for line in onDisk:
   diskL.append(line)
savedS = set(savedL)
diskS = set(diskL)
notSaved = diskS - savedS
notSavedF = open("",'w')
notSavedF.writelines(notSaved)
saved.close()
onDisk.close()
notSavedF.close()

[this has been edited once. it was 30 lines.]

So, none of you have questions for your government?

2007-04-01T14:57:43Z

Come on! I'm even doing all the work for you!

So far, I've got:

climate change, economic security and power,
disclosure of government policy,
the governments behavior during parliament, and
open formats in government offices.

What else do we need to ask?

Oink

2007-03-29T17:16:18Z

"You can put lipstick on a pig, but it's still a pig, and you're still not going to want to kiss it." -TWX

"And what I've found with Windows is that people take that pig and say "Well, it works well enough, and all the clothes I bought for it fit, so I don't think I could switch over to a real woman. Besides, I hear that real women are very expensive." -Paul Rothrock

I guess I should mention something about my move to Linux. It hasn't been without plenty of teething problems, and some of them persist, for example, I can't find the config file for ksh to get it to use readline. For some reason, scim isn't really working either. I've downloaded the Chinese and pinyin input methods, but I can't even get the X applet to show. On the terminal, everything looks peachy, but I just can't get it to work in X- so I can't get it to work in OpenOffice.org, which Jane needs for business.

Still, there is only one thing that is sub-standard about the Linux installation I'm using. I've got all of my data on an NTFS [windows] drive. NTFS doesn't want to play nice for some reason. I can understand it not supporting Linux-style access control lists, and so I could handle having to be in root to play with all my old data. However, I can't save back to that disk from Linux. Maybe this isn't much of a problem, at least it won't be when Linux ends up on my laptop, because it's quite likely to be the only operating system on there.

Along that theme, I decided today that I don't want another laptop. My laptop might be a tad antiquated, with its 256 mb of ram and 500mhz processor, but modern machines don't seem to be running that much faster. $5000 won't buy you more than dualcore 2.33 ghz laptops, and as far as I can tell, they don't have x86 virtualisation extensions. Maybe I'd be better off just buying a gig of ram and a nice big hard disk. The cold hard reality is, 2ghz dualcore will still take half a day to do any decent mathematical calculation, and my laptop is more than adequate for work, music composition, programming, LaTeX, and the completion of The Shwa Project [assuming I can get magic VLSI working, I'm getting a weird but mistaken error concerning gcc].

At this rate, I guess I'll be waiting for a terrascale system before I look back into buying computers. Hopefully we'll have simple Shwa machines put together by then. I was looking through OpenGL today and decided I might be able to do Albert-based OpenGL hardware, the 45-nm possibility suggests we could push >80 billion polygons / sec. With a dedicated design we could see a form factor of what, four millimeter square? hmm...

OK now I'm just talking crap. Anyway, I think I'm going to write a letter to Ms Julia Gillard. Not that I have much faith in government at all, but I can't rest without feeling I've done something to stress the things this country needs.

Like how coal miners jobs are bought up every time anyone makes a mention of alternative energy. Surely, if we are producing the same amounts of everything while less people have jobs, then we are more efficient as a country. Those people can find new positions. If it gets too bad, in the sense that benefits are not being distributed evenly, it will be left to government to suggest some sort of communist scheme whereby the people are able to enjoy the fruits of their labor. I don't see how upsetting those positions down the track instead of now [it will need to happen eventually!] is any better.

I'd also like to see some policy. You know, rather than marketing hype and government bitching. I'd like to see what Labor plan to do, and I think the rest of the public does too. We don't care about where our PM grew up, he's not a pop star and never will be [granted, Beasley was the reason I didn't vote Labor last election]. I want to know where we will be moving forward and where we will be moving backward. Will any remaining government assets be secure, or will there be more things sold? Where will our money be going? What will the government do to address the aging population? And speaking of openness, what about the recent push for governments to use all Free software? Can we see similar laws passed here?

Now I want to know what YOU think. What really gets your goat? What do you want to see from our next government? What do you want changed? What do you want looked at? What do you want gone? What are your concerns? Why? If you're timely, I might include something in my letter :)

blee

2007-03-08T17:00:50Z

just some photos.

topology_n_lust @ 2007-02-24T17:16:00

2007-02-24T17:55:42Z

Just felt like jumping up and down briefly; I spent all morning generalising the full-adder I've always used, to handle subtractions too, and realised I could cut the number of gates from twelve to seven using x-or's cleverly. this suggests I might be able to subtract n-bit mantissas using $(8+\floor(\log_{2}(n)))*n$ gates. For n = 255 [read: shwa], that's some 3825 gates, some 22k transistors. I also think I've found a way to make flash ram more reliable, and faster, but it depends upon the open-circuit resistance of enhancement FETs. Think this'd make a decent first post in my tech journal? Not quite- there's much more to come.

Today Sam and I played chess, which was really cool. I got beaten twice; I really need more practice. Anyone feel like a game?

Turbo-Encabulator

2007-02-20T05:37:57Z

This is one of the funniest things I've ever read. I recommend you check out the GE datasheet too, especially the case material section. There are a couple of hillarious videos documenting the use and maintenance of Turbo-Encabulators on youtube [Here's one with some extra bits] that you should check out too.

I have a point to make about the FAQs though, the guy suggests a Laplace transformer over a Fourier transformer for powering a Vintage TE. The only thing is, Laplace transformers are known to decay very quickly. Traditionally, Mellin transformers have been used on older Turbo-Encabulators, and although there's a little mucking around with Mellin transformers, they integrate rather easily, and they capture the lower and the upper half complex plane in one pass, hence the use of helically polarized inputs. The switch to a simpler charge bifurcation stage in the 40's allowed for fewer grooves in the lunar waneshaft, and clearly, greatly simplified the use of the transcendental dadoscope. Hankel transformers were also specifically designed for powering tremie pipes, though they were awkward when the machine was used with a drawn reciprocating dingle-arm, with the exception in insulated nofer trunion machinery.

I guess now you know what it's like trying to find something out from the technical literature at work. Lately, I've been doing some stuff on Moment of Inertia. Fun fun.

A True Fiend? >:-)

2007-02-19T10:09:29Z

Because Robert, it's contagious...

Create your own Friend Test here

Got a new Blog for Mathgasms and such. Will post a link soon!