2009-11-01 11:05 - Gaming
I've more-or-less recently played through a bunch of games, without abusing them as an opportunity to fill my blog. Roundup!
Scribblenauts (DS)
This was heavily hyped. I was excited to see it. It was very underwhelming. The levels are too rigid to make truly imaginative solutions. And the control scheme is horrible. You're required to move objects around all the time, but miss them by just a hair and instead your character will run towards them, often to his doom. Didn't even finish this one, which is very out of the ordinary for me.
Condemned 2 (PS3)
Not bad. Not great, either. Randomly introduced long and slow sequences, or ultra-fast play it three times before you figure out what's killing you style bits. The "investigation" tasks were too frequently a random guess. I think, overall, this game tried to be "too much", with fighting, shooting, puzzle solving, racing, and a bit more scattered all throughout.
Chrono Trigger (DS)
I'm pretty darn sure I played this one before. But I'm an RPG fan, and this one is good enough that I played it right through again.
Professor Layton and The Diabolical Box (DS)
A solid sequel. A collection of good puzzles, and just enough story to tie them together. A few too many of the shuffle-the-pieces-around for my taste, but very worth playing.
2009-10-28 15:55 - Links
When you show us a bottle of liquid, we can’t tell if it’s a sports drink or liquid explosives without doing a time consuming test on it. We’re developing the proper technology to allow us to expedite the screening of all liquids, but in the meantime, to screen everybody’s various types of liquids over 3.4 oz. would cause gridlock at the checkpoints.
Shenanigans. If I am carrying a soda/water bottle/any other drink, there is a fast, easy, safe, and cheap test to see whether it is actually a potable drink, or a dangerous explosive chemical. I can drink it (or at least some of it).
2009-10-21 15:25 - Linux
Unix has a long-steeped tradition and philosophy which I normally agree with: create small tools that work together to solve bigger problems. It works really well at the command line, with classic tools. Chaining pipes together can get quite a few things done. It's totally failed in the GUI world, however.
I'm used to how all my programs act on Windows. If I launch Firefox (or Thunderbird, or Pidgin, or EditPlus, or Photoshop, or Volume Control or ... the list goes on) and put it in the top-right corner of the screen, close it, and reopen it: it comes back in the top-right corner of the screen, where I put it! I'm extremely used to working this way. It follows the desktop metaphor very well, I put particular things in particular places, and I'm very used to looking at that particular place to find that thing. (Aside: It's one of the things I really dislike about the 30" monitor I have at work now. There's a lot of space, but I'm used to the dual-monitor setup which divides said space into separate places. But I digress.)
It seems that in Windows land, every single program implements the "remember my size and position" logic, and restores said settings next time. The exceptions, which do not do this, get launched in a "cascade" arrangement: starting from the top left, each is down and to the right by a little bit, about (exactly?) the height of the title bar. The command prompt is a classic example of this. It's not useful. But, since Windows offers nothing better in terms of window placement, every (okay, almost every) piece of software solves this problem, and basically all in the same way: launch in the same place that it was last time.
One way to make this clear is to look at Pidgin. By default on Windows, the Pidgin "Buddy List" window, and each "conversation" window launches in the same position it was in last time. On almost every computer I use, for me that is the top-right and bottom-right corner respectively. I usually keep at least a bit of those corners visible at all times, and can easily switch to those windows with the mouse (see Fitt's Law: stuff on the edges, and especially in the corners (both edges) is really easy to get to with the mouse), and can see when something is going on in that window (or if it appears suddenly). And on Windows, this works. On Linux? Not so much.
See, Linux follows the Unix tradition, and things are often built in small cooperating pieces. So the GUI system as a whole delegates to the window manager ("software that controls the placement and appearance of windows within a windowing system in a graphical user interface") all the work of placing, sizing, and interacting with the actual windows holding the programs. As a result people who write Linux software (even cross-platform capable software) seem to think along the lines of, "Pidgin doesn't place the window, it is up to your window manager to decide the position."
But the window manager manages the windows horribly! First, I have to install a package that doesn't come by default (CompizConfig Settings Manager, aka ccsm) to get to the "Place Windows" setting. The choices are cascade, centered, smart, maximized, and random. What do these settings do? Starting from an empty desktop, and opening a few different programs, I see:
- Cascade
- Starts at an inconsistent position, near but not at the top-left corner. Following windows go in a column downards, the left edges aligned, until the vertical space is consumed. The next columns start at the same vertical position as the first window, unless this window is taller than the first window in the previous column, in which case it starts at the same vertical position as a lower window in the previous column. (The top-left corner of the new window starts at the bottom-left of a previous window if possible, otherwise at the top-right corner. Once neither are possible, the scheme starts over at the top-left of the whole screen. Each new window goes in the "first" available space.)
- Centered
- Every single window starts perfectly centered in the screen, all on top of each other.
- Smart
- Each window goes as far towards the top-left corner as it can, without covering any other windows. If there is no empty space big enough to hold it, it goes wherever it covers the most empty space (the desktop) and the least other windows possible.
- Maximized
- Every window takes up the entire screen.
- Random
- Just what it says on the tin. A surprise every time you open the window! Where will it go this time?!
So what are the merits of these schemes? They all stink! Well, I can sorta see the value of maximized on a small screen (netbook). But there are still certain apps (calculator, IM, etc) that have a smaller window and are just fine that way. Cascade is very difficult to predict (look at the description!) and is generally useless unless you're starting a whole bunch of the same program, in which case it will set up a nice grid. But that grid will be offset from both the top and side of the screen, wasting space.
Centered is pointless unless you enjoy spending your time rearranging your windows so that you can see them. Random is just as pointless for (nearly) the same reason. The chance that the window goes where you want it is virtually zero.
Smart is what is left. In certain, rare, cases, it's very useful! When I launch that one-off program, I probably do want it to not overlap anything. I might be firing up a calculator, and want to still be able to read the web page/document/etc that I'm getting the numbers from. But it fails miserably for the programs that I use every single day, and probably multiple times in every day. I want those to be in a particular place, and I want it every time.
As mentioned, it's the viewpoint of (some!) programmers that the window manager should be managing the windows, so i.e. Pidgin leaves me stuck. There is a workaround in the "Place Windows" plugin. But it reveals the reason that the program, and not the window manager, should be responsible. I can't choose carefully enough where which window I want to place. The choices are limited, most programs don't expose any distinguishing information, and even if they did, one big slider from -32767 to 32767 just doesn't get me where I want to go.
Why does it work on Windows? Because Windows doesn't even try. It doesn't try to stuff all this information into a narrow difficult API. Each program knows what windows it has, and how to identify each. And it stores four simple numbers (x, y, width, height) and uses them each time. So simple and easy that nearly every program does it. But that's too tough for Linux. It's even been pointed out, numerous times:
So where am I left? I managed to hack together some rough placement rules. Not many. Some programs (Firefox) get it right. Some programs get it wrong. Both my Prism apps (mail and calendar) start in the right place, except down and to the right by the exact space that the window border/title takes up, and must be moved each time. Eclipse starts in the right place, but is always six pixels too narrow, and nineteen too short. The "smart" placement mode almost gets the rest right. But there are still quite a few cases that I get surprised with where the window ends up, based on some funny coincidence of what screen space happened to already be used, and what desktop space happened to be peeking through at the time. Sigh.
So, to conclude: Where are we and why? Window placement is better on Windows than on Linux. Because the Unix philosophy, combined with the FOSS community, has let us down. The app developers say "the window manager should handle it" They even explicitly disable their window-position-remembering code that's already there, for Linux, because in theory some other layer handles the window management. The window manager developers say "the application should handle it" and "this should be done in user space" and the desktop environment developers say "this is impossible to do in user space". All the while, the users want it and are not getting it. Everyone along the chain claims someone else should do it, and closes the tickets as WONTFIX. And the users continue to suffer. Another item in the list of things that work fine on Windows, but are painful and/or annoying on Linux.
2009-10-18 10:10 - General
For as long as I knew about it, I've had a soft spot for The Far Side, an awesome one-panel cartoon that ran from 1980 to 1995. From 1995 through 2002 it was customary for me to receive an "Off the Wall" (aka desk) calendar as an xmas gift. Thanks to the hoarding impulses I've inherited from my mother, I still have them. A few are missing a month or two's worth of pages, but for the most part they're entirely intact.
My intent was always to create a digital archive of the images. Probably to use as a screen saver. Once, I even tried doing it by hand. Open photoshop, place the page on the flatbed, hit scan, wait, use the ruler tool to rotate so the top edge was straight, crop, save, repeat, repeat, repeat. I even got through most of one year. But it took far too long, and I bored of it quickly.
I've been putting some effort into cleaning house recently, and came across a couple of these calendars. Suddenly it hit me: I've got an auto-document-feeder on the scanner of my all-in-one printer, and I've got programming chops. I don't need to do this by hand!
The first, and surprisingly difficult, part was to find a program that would take advantage of the document feeder, and give me a pile of images to process. It was really easy to find commercial programs, but I didn't want to pay. Finally I discovered that IrfanView, a freeware image viewer also has a batch scanning feature, which suits my needs perfectly. Then I threw together a python script, which crops off the edges and transforms the scanned comic (and caption, if any) to be perfectly square, and crops with a consistently sized border.
It actually still takes a bit of work to babysit the scanner, making sure it feeds the small pages properly, and giving it batches that it can handle (not much more than 2 months' worth at a time). But it is mostly automatic, and I'm working through the sixth of the eight years worth I've got: the end is in sight.
Finally, I also found VisiPics, a convenient and free duplicate-image-detector, which works wonderfully for locating comics popular enough to make it into more than one calendar.
2009-10-14 10:36 - Linux
It's a very common paradigm for modern software to be able to open multiple files at once. It's nearly as common for software to present a "tabbed" interface for this, but this applies to any multi-document interface. In Windows, every single application (except for truly rare cases, I can only think of one I ever saw) uses Ctrl-Tab and Ctrl-Shift-Tab to shift to the next and the previous document, respectively. But, so far in Linux I've seen at least:
- Firefox, Chrome, Pidgin: Ctrl-(Shift-)Tab and Ctrl-PgUp/PgDown
- gedit: Ctrl-Alt-PgUp/PgDown
- Gnome Terminal: Ctrl-PgUp/PgDown
- Eclipse: Ctrl-PgUp/PgDown and Ctrl-(Shift-)F6
By far the worst thing is that for all these programs that have some other document navigation key, Ctrl-Tab continues to perform some other, significantly less useful, feature. Thus I'm always getting tripped up. Another way that the interface of Linux gets in the way: I have to think carefully about which program I happen to be using right now, in order to perform the same logical operation.
2009-10-11 20:18 - General
If I'm shopping for something significant (and if I am going to buy it, I'm going to spend time shopping for the best deal), chances are the method described below is how I do it.
I've used Yahoo! Pipes for a while. It's an awesome service. For a time, I used it to monitor various deals sites. I later added a "user input" option, to filter for particular items. Today, based off of an example I found, I just added a maximum price filter, too. Why? I use it to keep an eye on cheap PS3 games (and have certainly found a few worth getting, that way). But filtering for "PS3" finds as many systems as games. Limiting the price finds what I want.
Sound good? Open the (Input Any) Deals pipe, pick a price and a keyword, and subscribe in your favorite RSS reader. If I can come up with some ideas some time, I can tell you this is how I'll find xmas gifts.
2009-10-05 22:50 - General
Just outside of the subway stop near my house, there's a storefront, which was once a gym (if I remember correctly) then became a strange tiny Mazda dealership. Then, more recently, a Volkswagen dealership. They didn't do a very good job with the rebranding. The Mazda sign is still there, covered with a thin Volkswagen banner — at least, during the day when the original sign's lights are off.
Does it belong on The Fail Blog, or There I Fixed It?
2009-10-05 10:32 - Linux
In my continuing series of Linux rants, here I go on about fonts. Font rendering in Linux is horrible. It makes things especially difficult to read, or simply unpleasant to look at. For my first example:
The above image is a screenshot of the Optimal Use of Fonts on Linux guide, with the default font set to Verdana (one of the suggestions this guide makes). I've gone through it, and attempted to do what it says. But I still see issues like this: sometimes certain letters will become "bold" in the middle of a sentence, for no good reason. I know there's no markup in that HTML page that should make it show up bold, it's just some rendering engine that has a fault. Both the uppercase W and V letters look really wrong. It doesn't seem to handle those slanted lines well. (You can see a bit of the same in the M and y letters, and a teeny bit in the s, too.) Leaving Firefox at the default settings is a bit better — but probably just because Verdana isn't very heavily used, so I haven't seen it broken.
But there's much worse:
In Windows:
In Linux:
Both of the above screenshots come from an online article, which isn't important, except that this one seems to make use of one of the worst offenders: Helvetica. First, a very common complaint I have: Linux seems to screw up the kerning. See how especially the "e" is skewed to the left, also the "o", such that these letters are mashed up touching the pervious, with empty space to the right. Take a look at "delete" on the second line. The same goes for "That" a few times towards the end of the first paragraph: on Linux, it's all jumbly. It's also, overall, noticeably smaller than the same font on Windows. Result? The Windows screenshot is pleasant, the Linux one hard to read, even comparing screenshots on the same screen. (I suspect a high DPI on the monitor I have at work exacerbates some of my font-too-small issues, but the screenshots both on this screen show Windows results look better, on it.)
This even after doing everything I could to improve the situation. Installing the msttcorefonts (Microsoft Fonts) seems to have helped a few things (especially Arial, I believe). But plenty still looks bad.
2009-10-02 13:27 - Linux
Ubuntu is a bit famous for it's "Long Term Support" releases. That page says that "A new LTS version is released every 2 years. With the Long Term Support (LTS) version you get 3 years support on the desktop ..." and "The two LTS releases so far are DapperDrake and HardyHeron." But it also says that they will "Avoid structural changes as far as possible, such as ... lots of library transitions." At Google, my desktop is based on Hardy, which was released on April 24, 2008, or just over a year ago. So it should have received updates now, and for another year and a half or so, right? Wrong.
I've been searching for a Windows replacement for (among other things) Launchy on Windows. The best I found was Gnome Do. After a short while, I tried to work some of its advanced settings. I saw plenty of documentation online about the Preferences dialog — but I couldn't find it anywhere! Well, the reason is that the last version in Hardy's repositories is 0.4.0.1, but that was released back in March of 2008, before Hardy's first release. And it's never been updated. Do is up to version 0.8, with a lot of new features — including the preferences dialog!
I tried upgrading myself. First I looked for all the DEB packages I need (and don't I remember this sort of crap, scouring for RPMs for RedHat back in the day). I didn't get far, resolving the entire dependency tree one at a time, by hand, before I hit on something I couldn't find the DEB for. So I tried building from source, in a similarly frustrating manner. Somehow I found the Badgerports repository, where someone has thankfully taken the effort to port modern packages into the "Supported" version of Ubuntu I'm stuck with. After adding that repository, I could actually install a non-ancient, featureful, version of Gnome Do!
(I'm still not quite satisfied with it as a Launchy replacement, it's harder to customize. But it's pretty good. I miss being able to just double-click the EXE and get whatever software I want.)
