discovered what it was. There's something different about the way
Firefox renders fonts for webpages and the way all my other (GTK+ and
Qt) applications render fonts. It's especially strange, since the
buttons and other form widgets render fonts like the rest of the GUI,
but the web page text is different somehow, and thus seems a bit
off. I actually discovered this because the laptop that I've been provided by
work has an insane native resolution of 1920x1200 on a 15" screen. Thus,
in order to actually be able to read all my fonts, I've raised the DPI
to 120. This works great in everything, except Firefox, which
stubbornly refuses to use the systemwide DPI setting for anything other
than UI elements. This wouldn't be too bad, because I could just zoom
the pages or raise the font size for webpages, but since form elements
use the system fonts/dpi, they are a completely different size font than
the rest of the webpages, causing no end of rendering weirdness.
Searching for ways to solve this problem was fruitless, since it seems
that there is no way to set the DPI for the web page text, and trying to
customize fontsizes to match everything else via CSS seems to be an
exercise in futility (and pain). In desperation, I installed Konqueror, and found that it properly
respected my DPI settings, but didn't quite feel enough like a dedicated
web browser for my tastes. So I remembered that Opera exists, and I've
installed that. So far, it's as polished and fast as I remember, and it
seems to work on more sites than it used to. They also removed all the
ads, so I'm a happy camper so far. It's amazing what consistent text
rendering across the entire UI can do for a person. (As an aside, I
think this may be why Evolution feels better than Thunderbird for
reading mail, despite Evolution's somewhat fragile network layer and
lack of proxy support) Anyways, we'll see how this Opera experiment holds out. We'll see how
long it is until I miss the del.icio.us plugin and Firebug.
Note: This was originally written at least a month ago, and somehow, It never got posted. I started up Evolution this morning after not using it for a while, and apparently it had this unsent message in it's queue. I guess my network setup wasn't working very well at all when I wrote this.
I've generally been annoyed that Network Manager in Ubuntu doesn't give you the option to save a dynamic DNS configuration when you save a network configuration that uses DHCP. Also, I'm unhappy that Network Manager doesn't save the Gnome proxy configuration as part of the configuration. But these are minor gripes. I have recently found out that while Network Manager will let me configure my wireless for a static IP, the resulting configuration will not actually work. At all. I can get associated with the access point, and the routes and other network configuration looks fine to me, but I simply cannot send or receive anything. (And in troubleshooting this, I found that Network Manager doesn't save your WPA password between invocations (when you have a static setup), so if you do something like change the configured IP address, you need to remember to also re-enter the WPA password, otherwise, it will happily forget it.)Anyways, I ended up having to figure out wpa_supplicant (which is surprisingly straightforward), and reconnect to the access point manually. Now things work, but with the general amount of polish Ubuntu has been showing lately, I really didn't expect to find something this integral to the system as this unfinished. It just makes me sad. I've seen some reports that Network Manager 0.7 fixes the static IP issues, and that's coming in Intrepid Ibex. Anyways, that's enough pointless complaining. I'm going to do something productive. And perhaps after I move, I'll find enough free time to see if I can help fix some of these problems with Network Manager, because it really is my biggest pain point in Ubuntu these days.
There are bad design decisions, and there are horrible design decisions. The exception class Timeout::Error in Ruby's stdlib is one of those horrible things.Ok, let me start from the beginning. In Ruby, the standard way to catch any recoverable error in your program is to catch StandardError. Anything that doesn't inherit from StandardError is generally the type of thing that you want to have crash your program. Now, Ruby has a library for having an exception thrown after a timeout. (We'll ignore for a second that this library is implemented in a provably unsafe manner as far as threads are concerned.) You just call the timeout method, and pass a number of seconds and a block. If the block doesn't terminate after the number to seconds, it is forcibly terminated, and the Timeout::Error exception is called. This would be all hunky dory except that Timeout::Error isn't a subclass of StandardError, it's a subclass of Interrupt. "So," one may say, "why even use this thing at all?" Well, because you kinda have to. Want to do something with HTTP? Well, the standard HTTP library happily uses the timeout library to do it's timeouts. And as a result, so does pretty much every other library that has to do something with HTTP. So you're going along in your nice, happy application, when you suddenly find out that even though you're rescuing StandardError so that a network failure doesn't bring down your application, well, that isn't quite true: the minute a network timeout happens, your entire application comes crashing down. The only answer I can find on the internet is that one should always just rescue Timeout::Error when doing anything that might call the timeout library. (rant from a developer that found this way before I did) What I don't get is how this is still in the standard library. Why not do the obvious thing and just make Timeout::Error a subclass of StandardError? (Of course, the aforementioned problems with the timeout library should be fixed, but I can see that taking a long time) While I forsook Perl in favor of Ruby because of Perl's many gotchas, at least Perl's evil tends to be in my code, not hidden deep in a library that I'm using. And I still miss Perl's copious, usually well written documentation. (Even CPAN modules tend to be rather well documented) Sometimes, the land of Python seems like it might be nice... (The grass is always greener on the other side, and there is always something hidden in the grass waiting to bite you.) (I also have a whole rant about the stupidity of Java's properties object inheriting from Hashtable, but that's a lot less insidious.) (Yet Another Parenthesized note: maybe I just need to go back to my warm, fuzzy, Perl security blanket.)