Dana Boyd’s recent post brings up a few good points, but one that particularly resonates is that Facebook’s interface did not make it clear what you are giving away by using the site.

I think that they way to address this is to have a tabbed screen preview of your profile and activities, where each tab is a sharing constellation – 3rd party API users, friends, friends-of-friends, everyone – and you could tune your profile by excluding classes of information to those constellations, and see who comprises those groups at the same time. You could click on the friends-of-friends tab and see, perhaps to your horror, what you are sharing with your mom’s coworkers, and who they are in a neat, sortable list.

I would still be a Facebook user if I could click on the API tab and turn it off, click on the Everyone tab and turn off everything but my name and profile picture, etc. But I can’t and because I can’t see what I’m sharing, Facebook seems to think I don’t care, and they’re wrong.

So I go a-hunting in man xinput and find that, if the name of the device has changed, I will find it by using xinput list, which give me this:

? Virtual core pointer                    	id=2	[master pointer  (3)]
?   ? Virtual core XTEST pointer              	id=4	[slave  pointer  (2)]
?   ? Logitech USB Trackball                  	id=8	[slave  pointer  (2)]
?   ? Macintosh mouse button emulation        	id=10	[slave  pointer  (2)]
? Virtual core keyboard                   	id=3	[master keyboard (2)]
    ? Virtual core XTEST keyboard             	id=5	[slave  keyboard (3)]
    ? Power Button                            	id=6	[slave  keyboard (3)]
    ? Power Button                            	id=7	[slave  keyboard (3)]
    ? AT Translated Set 2 keyboard            	id=9	[slave  keyboard (3)]

It looks to me like it’s still called a “Logitech USB Trackball”. Long story short, I now have to identify this by id code, or drop the extra quotes. For posterity, here are the commands that work:

xinput set-int-prop 8 'Evdev Wheel Emulation Button' 8 8
xinput set-int-prop 8 'Evdev Wheel Emulation' 8 1

or

xinput set-int-prop 'Logitech USB Trackball' 'Evdev Wheel Emulation Button' 8 8
xinput set-int-prop 'Logitech USB Trackball' 'Evdev Wheel Emulation Button' 8 1

Yes, these changes are a pain in the ass.

I filed a bug report, and was informed that it’s not a bug. I disagree, in that if you break things in an upgrade, it is a bug.

Never-the-less, I had to figure out how to get my EmulateWheel behaviour back since the upgrade.

The evdev man page is nigh on useless, but this blog entry helped a lot.  It has lead to the following as my /etc/hal/fdi/policy/marble_mouse.fdi:

<match key="info.product" string="Logitech USB Trackball">
 <merge key="input.x11_options.EmulateWheel" type="string">true</merge>
 <merge key="input.x11_options.EmulateWheelButton" type="string">8</merge>
 <merge key="input.x11_options.ZAxsisMapping" type="string">4 5</merge>
</match>

I got the info.product part by looking at /dev/input/by-id/, and
guessing. I logged out, and restarted hal for good measure. One or the
other may not be necessary.

# /etc/X11/xorg.conf (xorg X Window System server configuration file)
#
# This file was generated by dexconf, the Debian X Configuration tool, using
# values from the debconf database.
#
# Edit this file with caution, and see the /etc/X11/xorg.conf manual page.
# (Type "man /etc/X11/xorg.conf" at the shell prompt.)
#
# This file is automatically updated on xserver-xorg package upgrades *only*
# if it has not been modified since the last upgrade of the xserver-xorg
# package.
#
# If you have edited this file but would like it to be automatically updated
# again, run the following command:
#   sudo dpkg-reconfigure -phigh xserver-xorg

Section "Files"
  FontPath  "unix/:7101"
  FontPath  "/usr/share/fonts/truetype"
  FontPath  "/usr/share/fonts/X11/misc"
  FontPath  "/usr/X11R6/lib/X11/fonts/misc"
  FontPath  "/usr/share/fonts/X11/cyrillic"
  FontPath  "/usr/X11R6/lib/X11/fonts/cyrillic"
  FontPath  "/usr/share/fonts/X11/100dpi/:unscaled"
  FontPath  "/usr/X11R6/lib/X11/fonts/100dpi/:unscaled"
  FontPath  "/usr/share/fonts/X11/75dpi/:unscaled"
  FontPath  "/usr/X11R6/lib/X11/fonts/75dpi/:unscaled"
  FontPath  "/usr/share/fonts/X11/Type1"
  FontPath  "/usr/X11R6/lib/X11/fonts/Type1"
  FontPath  "/usr/share/fonts/X11/100dpi"
  FontPath  "/usr/X11R6/lib/X11/fonts/100dpi"
  FontPath  "/usr/share/fonts/X11/75dpi"
  FontPath  "/usr/X11R6/lib/X11/fonts/75dpi"
  # path to defoma fonts
  FontPath  "/var/lib/defoma/x-ttcidfont-conf.d/dirs/TrueType"
EndSection

Section "Module"
  Load  "i2c"
  Load  "bitmap"
  Load  "ddc"
  #Load "dri"
  Load  "extmod"
  Load  "freetype"
  Load  "glx"
  Load  "int10"
  Load  "vbe"
EndSection

Section "InputDevice"
  Identifier  "Generic Keyboard"
  Driver    "kbd"
  Option    "CoreKeyboard"
  Option    "XkbRules"  "xorg"
  Option    "XkbModel"  "pc104"
  Option    "XkbLayout" "us"
EndSection

Section "InputDevice"
  Identifier  "Configured Mouse"
  Driver    "mouse"
  Option    "CorePointer"
  Option    "Device"    "/dev/input/mice"
  Option    "Protocol"    "Auto"
  Option    "Buttons" "4"
  Option    "EmulateWheel" "true"
  Option    "EmulateWheelButton" "8"  #Option   "Emulate3Buttons" "true"
EndSection

Section "Device"
  Identifier  "nVidia Corporation C51 [GeForce 6150 LE]"
  Driver "nvidia"
  #Driver   "nv"
  BusID   "PCI:0:5:0"
  Option "SWcursor"
EndSection

Section "Monitor"
  Identifier  "Generic Monitor"
  Option    "DPMS"
  HorizSync 28-84
  VertRefresh 43-60
EndSection

Section "Screen"
  Identifier  "Default Screen"
  Device    "nVidia Corporation C51 [GeForce 6150 LE]"
  Monitor   "Generic Monitor"
  DefaultDepth  24
  SubSection "Display"
    Depth   1
    Modes   "1680x1050" "1024x768" "800x600" "640x480"
  EndSubSection
  SubSection "Display"
    Depth   4
    Modes   "1680x1050" "1024x768" "800x600" "640x480"
  EndSubSection
  SubSection "Display"
    Depth   8
    Modes   "1680x1050" "1024x768" "800x600" "640x480"
  EndSubSection
  SubSection "Display"
    Depth   15
    Modes   "1680x1050" "1024x768" "800x600" "640x480"
  EndSubSection
  SubSection "Display"
    Depth   16
    Modes   "1680x1050" "1024x768" "800x600" "640x480"
  EndSubSection
  SubSection "Display"
    Depth   24
    Modes   "1680x1050" "1024x768" "800x600" "640x480"
  EndSubSection
EndSection

Section "ServerLayout"
  Identifier  "Default Layout"
  Screen    "Default Screen"
  InputDevice "Generic Keyboard"
  InputDevice "Configured Mouse"
EndSection

Section "DRI"
  Mode  0666
EndSection

1. Put the TTF files (there’s usually three) in their own subdirectory of /usr/share/fonts/truetype.
2. As root, run these two commands in the new subdirectory: mkfontscale and mkfontdir
3. Check to see if the font is installed: run xlsfonts and see if your font is there
4. If you don’t see your font, start Googling, but perhaps start here.

1. On system 1, make a file with your installed/removed packages: dpkg --get-selections > packages.txt
2. Get a base install done on system 2, and set /etc/apt/sources.list to be the same as on system 1
3. Get packages.txt onto system 2
4. Run dpkg --set-selections < packages.txt on system 2
5. Run dselect install on system 2
6. Optional: Dance

After much thrashing about, I learned that when I hit C-a z (Control a, z) I put the session in the background, and it can be returned with the command fg. Similarly, if I hit C-a s I have locked the session to input, and to return it I need to hit C-a q. Good things to know.

I was poking around with rename, but I wasn’t finding a good way to do this. Thankfully, a poster on TLUG showed me how to use sprintf to get the results I wanted, like so:

rename -n 's/(\d+)/sprintf("%04d", $1)/e' *.png

Very neat.

<script type="text/javascript">

function gotodate() {
  var baseurl = "http://example.com/index.php/"

  var y = document.getElementById("year").value;
  var m = document.getElementById("month").value;
  var d = document.getElementById("day").value;
  //debug alert(y + m + d);

  var newloc = baseurl + y + "/" + m + "/" + d
  window.location = newloc
}

<form>
  <table>
    <tr>
      <td>Year:</td>
      <td>Month:</td>
      <td>Day:</td>
      <td> </td>
    </tr>
    <tr>
      <td><input type="text" id="year" size="4" maxlength="4" /></td>
      <td><input type="text" id="month" size="2" maxlength="2" /></td>
      <td><input type="text" id="day" size="2" maxlength="2" /></td>
      <td><input type="button" value="Go" onclick="javascript:gotodate();" /></td>
    </tr>
</table>
</form>

There is no good way to know what you know – there is no feasible method to list your mind’s contents, and, knowing even obliquely the nature of the mind, it could easily spool off infinitely. The best you can do is to hope that others are able to figure out the easy drudgery of your job and be hopelessly in your debt for the interesting things that you do, so you are appreciated without having to be buried in minutia when you return.

:0:
* ^(From|To|Cc).*\@listdomain.eg
  {
    :0hf
    | /usr/bin/formail -A "Reply-To: listaddress@listdomain.eg"
    :0
    folderforlistmail
  }

I much prefer to hack the client-side then try to change policy on the server-side. I am almost universally less resistant to change than any organization.

After so poking about, I only used a subset of those instructions. I added this to my .screenrc:

attrcolor b ".I"
termcapinfo xterm 'Co#256:AB=\E[48;5;%dm:AF=\E[38;5;%dm'
defbce "on"

My .vimrc got updated with this:

set t_Co=256

It did a great job, and now everything looks as it should.

Here’s how I use it; I have four primary email accounts, each with their own .muttrc. I open a screen session for each account, plus one session as a scratch pad, for the variety of non-mail activities that I do each day. (If you’re thinking that one session for all non-mail activities is too few, you’d be right most of the time, but I have a solution for that too.) I start all of these sessions by calling screen with a special config file, which I call .myscreenrc, to separate it from the regular .screenrc. It looks like this:

autodetach on
shell -$SHELL
screen -t scratch
screen -t uoft u
screen -t nerd n
screen -t witteman w
screen -t woolgathering wg

The -t option givens me a title for each session within screen – which helps me keep everything straight. I call this with a line in my .xinitrc, like so:

sleep 1 && urxvt -geometry 80x56+0+0 -e screen -c .myscreenrc &

The delay helps it come up after my backdrop is drawn, so I don’t have a blank behind my transparent terminal.

All of this is now nicely set up, and I am in mutt (the last email account). I hop between instances in three ways: Ctrl-a " for a list of sessions to scroll through, Ctrl-a ' and the number of the session I want to be in or Ctrl-a n or Ctrl-a p for the next and previous session. If I need a fresh session, it’s as easy as Ctrl-a c.

There are a couple of problems with this setup. The mutt sessions only lasts as long as mutt is open, so if my fingers, from long training, close mutt, then I lose that session as well, and I have to start it afresh. The solution, rather than training my fingers, was to prevent mutt from closing. I remapped q in the browser and index modes to be the equivalent of c?, which means that I can only close mutt from the pager by hitting x. Fine by me, and it lets me have long-lived sessions.

Now, whether I am sitting at my machine or ssh-ing in from work, I have all the same things at my disposal, and I can comfortably leave aspects of my work open when changing locations. If I have left my sessions running at home, I simply call screen -x when at work, and I am in my familiar environment. If I have detached at home then screen -r is the right call.

However, if I have detached at home and stopped my X session, I run into the other problem. When I started screen via my .xinitrc I also set the $DISPLAY variable in each session. As such, when I am logged in remotely without X running, certain programs throw errors because I am not able to reach the X session. The sidestep for this is to unset DISPLAY if I am going to be using programs that care about that variable, most notably vim.

It sounds like work, but it is actually extremely sweet, and really easy.

What’s troublesome is that there is no equivalent in windowing environments. It would be hugely useful to be able to detach from an X session – even if it was just locally – so someone else could log in and use the machine without me having to close everything down. The Microsoft Remote Desktop makes some attempt to do this, but it is clunky and does not behave consistently. Ideally, when you start the machine it should launch a server process which in turn launches a login screen (or prompt) as a virtual machine or client instance, and whenever you wanted you could detach that instance from your display and input hardware, and then launch another instance. The standard should be a thin-client system, so you can do more with the same machine. This would be especially useful in the home, where one machine could have several heads, and anyone could use “their” computer from any head.

On Linux, I used the xmodmap command (because I spend all my time in X – the console is not necessary) in my .xinitrc file like so:

xmodmap -e "remove lock = Caps_Lock"
xmodmap -e "keycode 66 = Escape"

Because I am shackled to a Windows box day after day, I also wanted to make this work in that environment, and luckily, I can. It just took a brave bit of registry editing, like so:

REGEDIT4
[HKEY_LOCAL_MACHINESYSTEMCurrentControlSetControlKeyboard Layout]
"Scancode Map"=hex:00,00,00,00,00,00,00,00,02,00,00,00,01,00,3a,00,00,00,00,00

It makes a huge difference to not have to reach for the Esc key all the time, but it makes it so easy that I keep shifting into command mode while typing in Word. A small price, especially since I realize I’m happiest writing in vim anyway.

One trick that I think is just the bee’s knees is *. If you are on a word in command mode, hit * and you will search for that word. Even better though, is the gd trick. When the cursor is on a variable name, typing gd will take you to the declaration of that variable.

The last thing that I think is awesome is "repeat the last change". If I am doing something repetitive, like creating an associative array of static values, I can type the first declaration and then start hitting . in command mode until I have enough declarations for the whole array. Very sweet.

What I discovered is that no one is willing to do this without at least a hundred bucks, which is idiotic. I learned from my research that the CMOS battery is encased in packaging, is difficult to remove on many ThinkPad models, and must be purchased as a specialty item via eBay or IBM. The technicians I spoke to where not able to order it until they had the laptop in their hands, because apparently there are different batteries for different 600X machines. I now believe that this is bullshit, and that they are simply too eager for my money to actually provide advice.

Rather than submit to the whims of the extortionists, I started removing screws and panels from my ThinkPad in search of the mythical battery. I found it in the RAM access hatch (I only needed to remove one screw) which is above (below?) the removable drive chassis (CDROM, DVDROM, etc.). It connects to the motherboard via two leads going into a specialty plug, and those leads connect to the battery via simple paddles. The battery and the paddles are held together by a short length of heat-shrink tubing. This sounds way more complicated than it is. The battery cover has the battery type printed on it, CR2025, which is a standard lithium battery I replaced at Blacks for $6. I detached the leads from the old battery, attached them to the new battery (being careful to replicate the polarity of the old one) with a length of electrical tape. I plugged it into the wee slot, replaced the panel, booted the machine and set the clock. It was essentially effortless, if you don’t count the visits to laptop repair shops and the subsequent anxiety.

Simple repairs can, in fact, be simple.

An interesting quote from the interview is this: "It used to be that there were lots of nuts with ideas in the world, but they were isolated nuts. In this day and age, there are no isolated nuts. You can always find in the world of the Internet 50 nuts just like you, and what do you call 50 nuts like you? A start-up company!"

If you use Windows, are you free to contact Microsoft and ask for help? No, not really – you maintain a busy in-house staff to take care of this for you, and to reinstall the operating system or reboot the workstation or server (!) as your primary troubleshooting methods. If there is a glitch in Word, can you do anything about it? No, you can’t. You can complain and whine, but you cannot fix it or get it fixed. Once a new version of Word is out, you can take your chances and buy a new license, and then hope that your problem got fixed, and that any new features don’t break your workflow, render your old documents unreadable or require a new training course. Most closed source software “support” is not supportive, and can only help you in narrow ways when the software is broken, but never when it’s wrong.

A common complaint of those confronted with the Open Source model is the claim that while you are permitted to change the source of your software, you lack the expertise to do so, which makes it a non-advantage. The counter to this is that, if you need a feature or a bug fix, you can simply sponsor a hack-a-thon to get it implemented, supporting the community and getting your needed fix far faster and more cheaply than hiring a consultant or having fully customized software written. There are a lot of changes required in the way that people think of software development in an Open Source world, but those changes will be easier to make once you shed your belief in the outmoded “support myth”.

I am not suggesting that Apple should have stuck to the PPC chips – they
have not been advancing and are not providing an advantage to the
company. You can be as big a fan as you like of the architecture, but
if no one can make it speed up, and no one can make it’s
"better" features actually perform better, then it is time to
move on. That said, going to Intel is idiotic. Intel has been getting
it’s butt kicked by AMD for a few years now, and they are still
providing us with 1970′s technology today. Most importantly, by going
to a commodity chipset, Apple has paved the way to having their precious
software work on cheap, generic machines, which will radically undercut
their hardware sales. Apple is a company that needs to focus – either
make hardware or software, but stop trying to do both as a seamless
"experience". Some suggest that assuming Apple doesn’t sell
OS X for generic hardware, it will get ported to generic hardware and
widely pirated. I think that this is true, and it will hurt both Apple
and Microsoft, because neither one will make money from it. Some
suggest that the lack of support will dissuade people from this, but
that is not th case for the home user at all (though it is a major
deterrent for business). How much support is provided by Apple or
Microsoft now? Sure, there’s someone you can call, but I’ll tell you to
reboot for free, and I can ignore you just as well as a big company.
User-level support is practically non-existent for any OS – you need a
community or a third-party service vendor to get any real help.

Perhaps the most important issue is control – with both Windows and OS X
you get very little control over the system you use, and if you exert what
little control you have, you frequently find that you have voided your
support contract. On Windows you have control over your hardware, but
if you change too much you’ll have to re-register your copy of Windows.
Also, if you want to change a behaviour or fix a bug you have to wait,
and to buy a new product sight-unseen in hope of fixing it. I am no
longer willing to do that. I no longer up proprietary software because
it denies me that ability to control my computer for no benefit.

The funny thing is, once I decided to use urxvt I ended up spending more time configuring it than I ever did getting Unicode (specifically UTF-8) set as my default character set. I’m pretty happy with it now though – you can check out my config file over in the configs section. The only problem left is that I don’t know how to type accented characters, but I figure I’ll figure it out eventually.

1. Make sure your wireless card is supported in your running kernel.
Contrary to the documentation provided by the Campus Wireless network,
any card supported by Linux and capable of WEP and 802.11b will work.

If your kernel does not support your wireless card, but it is supported
under Linux, you will need to recompile your kernel. This is beyond the
scope of these instructions, so get on the Internet and get some advice.

2. Note the output of iwconfig. It may look something like this:

lo        no wireless extensions.

eth0      IEEE 802.11-DS  ESSID:"UTORwin"  Nickname:"Prism  I"
          Mode:Managed  Frequency:2.412 GHz  Access Point: 00:01:F4:6B:0E:CC
          Bit Rate:11 Mb/s   Tx-Power=15 dBm   Sensitivity:1/3
          Retry min limit:8   RTS thr:off   Fragment thr:off
          Power Management:off
          Link Quality=46/92  Signal level=-78 dBm  Noise level=-149 dBm
          Rx invalid nwid:0  Rx invalid crypt:15  Rx invalid frag:2
          Tx excessive retries:2  Invalid misc:0   Missed beacon:0

The key fields are ESSID, Access Point, and Link Quality and the name of
the interface (eth0, eth1, etc.). If the Access Point field is
44:44:44:44:44:44 and the Link Quality is reasonably high, then you have
not been authenticated by the Access Point.

3. As root, log onto an access point with the following command:
iwconfig eth0 essid UTORwin enc s:UToronto1home
The only difference will be the interface name. You may wish to alias
this command to something convenient, or set it as the default for the
interface.

4. Log onto a web browser, and any page you receive will redirect you
to the Campus Wireless Network, where you will have to register the
MAC address of your wireless card and then authenticate yourself based
on your UTORID.

Getting a UTORID I leave as an exersize for the reader.

I did have a little trouble getting it working on Linux, but once I determined that the card was being mounted as /dev/sda1 (hurray for tail -f /var/log/messages) it was pretty easy. One other wrinkle was that the card is formatted as FAT16, which has an allocation-block limit of something silly, like 512 blocks in the root of the drive. I tried to fill the card with music, and got stopped at 90-odd files, about half-full, and I couldn’t figure it out. Eventually TLUG came to my rescue, and I learned that FAT16 breaks long filenames into blocks of a set size, so the 90-something files with long names became 512 full blocks. I created a directory on the card, dumped all the songs into it, and was able to fill the card. It seems that the limit is only for the root of the device. I’d love to use a better format, but I want it to work on any system I encounter – including the MP3 player, so I’m stuck for now. There is a 4Gb limit to FAT16 too, but I don’t think that’ll be a problem in the next year or two.

Okay, back to the kernel – I went through the process again, and I still had USB issues, but everything else worked. I got a bit of help from TLUG and discovered that my problems went away if I typed sudo modprobe ehci_hcd and sudo modprobe uhci_hcd, which are the modules for USB 2.0 and USB 1.1 respectively. After a bit more looking, I found that the reason I had some old, unused modules loading was that they were in /etc/modules. So, I extrapolated, put the USB modules in and commented the obsoleted ones out, and Bob’s your uncle, everything works.

cd /usr/src

tar xjfv linux.[version].tar.bz2

rm linux (assuming it is a link to your old kernel)

ln -s /usr/src/linux.[version] linux

cd /usr/src/linux

cp /usr/src/[oldkernel-source-directory]/.config .config

make oldconfig

make menuconfig

cd /usr/src/linux

make-kpkg clean

make-kpkg --append-to-version=.[a_unique_and_clever_name] kernel_image modules_image

cd /usr/src

dpkg -i kernel-image-[version].[a_unique_and_clever_name]_10.00.Custom386.deb

vi /etc/lilo.conf

lilo

What gave me troubles was using ALSA instead of OSS as the sound architecture. It worked, after following instructions to install a few things, if I was root. I don’t always want to be root though, and nothing I tried seemed to work. Finally, I used the brute force approach. I went into /dev and went to town, chmoding everything in sight, including mixer, sequencer, dsp, music and the snd directory, as well as the entire contents of /dev/snd. It works now though, so that’s something.

I tried aterm, because it had little in the way of dependent libraries, but it didn’t refresh when the backdrop image did, which I found annoying. So, I tried eterm (pronounced, inexplicably, Eterm) and kicked it around for a while. I think I like it, based on this theme:

<Eterm-0.9.2>
begin color
    foreground white
    background black
    cursor #ffff00
    cursor_text #880000
    pointer white
    video normal
end color

begin attributes
    geometry 80x56
end attributes

begin imageclasses
    path "/usr/share/Eterm/pix/"
    begin image
        type background
        mode trans allow trans auto
        state normal
        cmod image 100
    end image
end imageclasses

begin toggles
    map_alert on
    visual_bell on
    login_shell true
    scrollbar off
    utmp_logging on
    iconic false
    home_on_output 1
    home_on_input 1
    scrollbar_right true
    scrollbar_floating false
    borderless false
end toggles

begin keyboard
    smallfont_key LessThan
    bigfont_key GreaterThan
end keyboard

begin misc
    save_lines 1024
    cut_chars "t`"'&() *,;<=>?@[]{|}"
    border_width 0
end misc

It is really simple, but lets me see my pretty backdrop images even when the window is cluttered with terminals. I had to go into my .muttrc to change the “black” to “default”, but that was no hardship.

When I’m debugging code it is just magic to be able to comment out a block at a time to test various aspects of a program. For a long time now I’ve been referring to a copy-and-pasted section of the vim manual in a loose file, because I don’t use the features so often that it sticks in my mind yet. To avoid collecting these loose files, I’ll put the relevant section here, so I can find it.

Visual-block Insert *v_b_I* With a blockwise selection, I{string} will insert {string} at the start of block in every line of the block, provided that the line extends into the block. Thus lines that are short will remain unmodified. TABs are split to retain visual columns.

Visual-block Append*v_b_A* With a blockwise
selection, A{string} will append {string} to the end of block on every line of the
block. There is some differing behavior where the block RHS is not straight, due to different line lengths:

1. Block was created with $ In this case the string is appended to the end of each line. 2. Block was created with {move-around} In this case the string is appended to the end of the block on each line, and whitespace is inserted to pad to the end-of-block column.

Note: “I” and “A” behave differently for lines that don’t extend into the selected block. This was done intentionally, so that you can do it the way you want.

Now, well after I really could have used it, this article pops up. It is a useful primer, but it doesn’t really touch on the one things that will probably bite a lot of users in the ass – there are zillions of versions of LaTeX and its packages, and so you’ll have to experiment to find out what works on your system. Still, it’s a good document preparation system, and I’m glad to know something about how it works.

# A directory called /mnt/key
# These lines in /etc/fstab
none        /proc/bus/usb usbfs defaults            0   0
/dev/sda    /mnt/key    vfat    rw,user,noauto      0   0
# And these lines in my .bashrc
alias key="mount /dev/sda;ls /mnt/key"
alias unkey="umount /mnt/key"

It works pretty sweetly, and these storage devices are awesome. I’ve never gotten close to filling it up yet.

After reading this I decided to give subversion a try. It was pretty easy to set up, and it works insanely well with binary data, because it uses binary differencing on all files, making the distinction moot.

Before I got started with subversion though, I wanted to clear out all of the CVS administrative directories so I wasn’t versioning those as well. Well, since there is a CVS directory in every directory in the tree, I didn’t want to remove them all manually. Luckily, TLUG came to the rescue with this: find . -type d -name CVS | xargs rm -Rf. Very nice, and useful in a lot of ways. I’ll have to spend some time with man xargs though, because I think that I’ll be glad to know it better.

I’ve got the site somewhat in place, but I am still working out bugs. Hopefully I’ll get that all sorted this weekend. If you see something wonky, let me know.

I finally got Debian reinstalled, but with the flaky motherboard I needed something else. What I found was a new motherboard, processor, RAM and case for the same money as a lesser processor. It was significantly faster than what I was used to, though I will still have to look further into CPU-detection, because dmesg is not reporting its full clock speed, but the BIOS, I think, is.

The upshot is that I didn’t use tasksel and I didn’t use dselect to install non-basic packages, so my system is pretty lean. I am always pleased to not have printing daemons installed when I don’t need them. Now of course I want to export my installed packages to a file to keep safe, so I can do one fat apt-get at the beginning of any installation.

There is always lots to do.

Anyway, after some recent computer troubles, I was chatting with emma and she mentioned that she uses some command to list her installed packages, and from there can reinstall her system in one big chunk. To me, with two computers, this is a fabulous idea, because if I need to reinstall one of them, I can do it from the installed software image of the other. Very pleasing.

After some poking about I found these two commands, which I think are going to make my life a bunch easier: dpkg --list and apt-show-versions -u. The first is the one that tells me what all I’ve installed – not as easy a thing to remember, even on a stripped-down system, and it is likely to be more than you think. The second is quite handy if, like me, you don’t want to run apt-get upgrade blind. It should be good.

I’ve moved /home a few times, and I have a procedure for that, but I wasn’t sure you could do the same thing with /usr. I asked Arthur, and he pointed out something very important – the editiors are on /usr, and so I had to edit /etc/fstab before renaming /usr. Once I had the order worked out though, it went perfectly. Here is the procedure:

# Set up a partition table on the new hard drive
cfdisk /dev/hdb

# Get a filesystem on the partition
# you're going to use
mkfs.ext2 /dev/hdb1

# Mount the soon-to-/usr partition
mount /dev/hdb1 /mnt/tmp/

# Edit the filesystem table
vi /etc/fstab
# Add the new /usr to the table
/dev/hdb1   /usr    ext2    defaults   1   2

# Get into /usr and prepare to copy
cd /usr

# Single-user mode
# a bare root for low-level maintenance
init 1

# Copy everything into the new /usr
cp -ax * /mnt/tmp/

# Rename /usr so that you can mount
# the new /usr in the right place
mv /usr /usr.old

# Set up the mount point
mkdir /usr

# Cross your fingers
reboot

I was pretty happy with Blackbox, but I didn’t like some of the ways that things were configured, especially the fact that I had to elevate my priviledge to edit my menu or style file. Openbox uses locally specified XML files for configuration, and I am very pleased with how that system works. It is especially nice (though it would be good if it were clearly documented) that if you screw up your local settings you will get the default — but fully functional — settings and menu.

My wishlist for Openbox is for better documentation (so I’ve started to write it) and better management of where windows appear. I run two monitors, and the windows will spawn on the left edge of the right screen, which is not where I want them. I’d also like them to appear in columns rather than rows if I so choose. Still, it’s a nice window manager with active participation, and I’m glad to use it.

Most of the process works as expected, requires minimal kicking and has been pretty satisfying on the whole. Getting X working, is, as always, the thorniest challenge, but with known hardware I can now almost casually sacrifice the correct chickens and scatter the right entrails to get where I need to be.

The real bastard was getting my mail system set up – I wrestled with exim for a long while – Debian stable is ludicrously conservative, and I was for a long time worried about both exim and mutt being too many versions behind my config files and knowledge. After much gnashing of teeth, the installation of postfix and an upgrade of mutt (to 1.54) I still couldn’t get the MTA to honour my setting of a From: address. Finally, I looked at /etc/Muttrc (which I had looked for earlier in the process but didn’t find because I was expecting a directory and lower case) and found this little gem:

# don't generate a From header
unset use_from

Once I commented that out things worked just fine. I’m really pleased to have sorted that out, but I am also really peeved at the Debian dork who set that as off by default. All MUAs may suck, but every one I’ve ever seen sets From: correctly.

The opening chapters outline the basic concepts really clearly, and walk you through the short course in database design. If you are creating a small database with one of these tools (MySQL or mSQL) this book will get you started in fine style. The chapter on installing the tools is comprehensive and very well laid out. The basic style of the prose is easy to read and very comforting if you are new to the topic.

The big problem with the book is that it is organized with the information about the two tools explained in parallel. This is quite simple to understand and easy to read, but once you start trying to flip back through the book you realize that you get hopelessly snarled by the subtle differences between the two software packages. MySQL and mSQL are so close in function, and implement SQL so similarly, that it is easy to get confused by the tiny differences in syntax and structure that that these two systems demand.

Alarmingly often I find that I can confirm and sometimes diagnose a problem with my ISP’s mailserver before they can. I also find it terribly handy for sorting out wonky behavior on my web server, though I just use an O’Reilly book for that (HTTP Pocket Reference).

I think it would be pretty useful if everything could be in Unicode, from OSes to networking protocols, but that seems rather a long way off. Just the ability to hack in telnet in your own language would be pretty neat. As it is, you’ve got to do things by rote or learn English. I’m glad I don’t have to do that again.

A while back I read this article, which I find fascinating. It is a very intriguing idea for the laptop, since it boots up so much more often than the desktop. I haven’t had the time to look into it more, but I’d sure like to.

The book is essentially a "State of the Union" document that shows you what is implemented in the browsers, how it works, where the bugs are and what, if any, are the workarounds. In this capacity it is hugely useful, but it isn’t always easy to use as a reference. The layout of the book, into chapters based on the effects you wish to use, such as Fonts, Boxes and Borders, and Colors and Backgrounds(sic) is very helpful on the surface, but because the properties, selectors and elements are spread through the text there is no easy way to flip directly to the section you want. The appendix of CSS1 Properties is alphabetical only, so if you don’t know precisely what you’re looking for you’ll have to skim the relevant chapter.

The chapters themselves are often very long, with some examples having related code and some merely described, so it is quite tricky to figure out how to replicate an effect that you find in one of the many figures. In some cases the idealized behaviour is laboriously discussed, and the browser implementation is mentioned only briefly, making for a frustrating read. The prose itself is fairly transparent, and Meyer has peppered his examples with whimsy which eases the read considerably.

There are certain chapters of this book that are "must-read", and others that can safely be mined for tips and samples without being gone over in detail. Chapter 2, Selectors and Structure, and Chapter 3, Units and Values, are essential reading for anyone wishing to use CSS now, or in the future. These chapters are specific, focussed and full of priceless pearls that must have been discovered after hours or days of hair pulling and frustration. Similarly, Chapter 8, Visual Formatting, on the surface an academic interpretation of the CSS1 Specification, is the perfect background for understanding big chunks of CSS of both levels. The last Chapter, CSS in Action, is a good place to start if you are already somewhat familiar with CSS.

One thing the quickly becomes clear when reading this book is that Netscape is the ugly step-child when it comes to CSS. The winner of most of the early battles in the Browser Wars, Netscape is now relegated to flirting with the nurses and giving away half of every pair of shoes. Netscape’s support for even the rudiments of CSS1 are intermittent and riddled with bugs, requiring elaborate workarounds and even the abandonment of useful effects. There is hope that Netscape 6 will come to the rescue, but as feature is added to bug-eaten and crash-prone feature that hope dims. The fact that the smart money in the Netscape 6 release-date pool is on dates in the second half of 2001 is not encouraging.

Ultimately, Cascading Style Sheets: The Definitive Guide, is a good book to have read, especially if you intend to stay at the bleeding edge of web publication, but it is not a great book to have on the shelf. A more useable CSS reference can be found in Web Design in a Nutshell, by Jennifer Niederst or in Dynamic HTML: The Definitive Reference by Danny Goodman.