Archive for the ‘Programming’ Category
I’m now convinced that .NET’s support for localization is very poorly designed indeed. While on the surface, the concept of localization (also referred to as internationalization) is a noble effort to enhance or make possible the use of computers by speakers of foreign languages, if it gets in the way of development it can be a very disruptive and problematic tool.
Take the following error an early beta tester of OrangeNote supplied me with:
System.Windows.Markup.XamlParseException: Verdien System.Windows.Data.BindingExpression kan ikke tilordnes egenskapen Opacity for objektet System.Windows.Controls.Border. Invalid real literal ‘1.0’ Feil på objektet System.Windows.Data.Binding i kodefilen OrangeNote;component/viewers/debugterminal.xaml, linje 122, posisjon 5.
Now the first and most obvious problem is that the error is not in English! While it’s true that error messages presented to the user should most certainly be translated to their native language whenever possible, exception messages are not oriented towards end users. The purpose of an exception is to convey information back to the programmer that a particular problem has arisen in the code of a program, with the intention of providing useful information to addressing and correcting the problem.
For example, when translated, the above reads roughly like this:
System.Windows.Markup.XamlParseException: Value of System.Windows.Data.BindingExpression can not be assigned to attribute Opacity for object System.Windows.Controls.Border. Invalid real literal ‘1.0’ Error in object System.Windows.Data.Binding in file OrangeNote;component/viewers/debugterminal.xaml, line 122, position 5.
Is this really something the end user needs to know? If not, why go through the trouble of translating it in the first place? Perhaps Microsoft expects my users to fix my bugs for me. Furthermore, the platform doesn’t even appear to provide any way to explicitly obtain an English version of an exception, further increasing the absurdity of the matter.
But this is only the beginning.
If you take a closer look at the problem the exception is describing you might notice something strange: .NET is complaining that “1.0” is not a valid value for the Opacity property of a System.Windows.Controls.Border (which is a Double). Come again? 1.0 is a perfectly valid value for Opacity, unless you’re running your program on a locale that represents decimal values differently. That’s right, we can’t even program our app in English anymore! When simply changing the locale of your system affects what applications you can run on it, this is a clear indicator of a serious design problem.
Granted, this issue arises in some atypical scenarios involving dynamic parsing of lambda expressions, but the fact that it arises at all is the problem. What started out as a means to make applications more user friendly to foreign language speakers turned into a roadblock that can actually interfere with applications running at all.
In closing, my recommendations to Microsoft are as follows: First and foremost, eliminate the ridiculous and useless translating of exceptions into local dialects, and second, provide a more intuitive programming environment that takes into account programmer locale when running an application, not just user locale. These two things combined will make for a much more pleasant experience both for programmers and end users in multilingual scenarios.
See you next time.
Since the beginning of the FeedBeast project, I’ve been on something of a quest to find the perfect self-extracting executable maker, and I’m delighted to say I’ve finally found it! You may have already heard of it; it’s called WinRAR.
Windows software deployment is a complex process. Coming from the Mac and getting dumped into the deep end of Windows deployment, so to speak, it was quite a shock. There are a great many things you need to worry about just to get your software onto your end-user’s PC, and the headaches are multiplied if you’re deploying a .NET-based app like FeedBeast or OrangeNote.
One of the key concerns in such a deployment is making sure that the user’s computer has the .NET framework installed, before even beginning the installation of your product. This is done with the aid of what’s called a “bootstrapper”. [I'm not entirely sure where that term came from, but I reckon it has something to do with the idea of "strapping" something onto your program, which is exactly what a bootstrapper does.] It’s basically a second executable that sits alongside your installer (usually called “setup.exe”) that performs several checks to make sure your system has the necessary “prerequisites” for your program to run and/or install, optionally downloading and installing them if they are lacking.
The problem that arises from this configuration is we now have a whole directory of junk to distribute, not a nice clean singular installer file like we started with. In today’s web-deployed world, this is a real problem: users don’t want to have to download and extract a zip file just to gain access to the installer. This is where a self-extracting executable comes in.
A self-extracting executable bundles all of these pieces together into a zip file, but packages that zip file as an actual executable program, with enough logic built-in that it can extract its contents and launch the bootstrapper silently, so that it appears as though the self-extractor is the installer. This is the perception we want to give the user, because it makes things all that much simpler if they can just download a single executable file, run it, and install the program they’re trying to get.
However, it’s not always so easily achieved.
The thing is, while the idea of the self-extractor is simple, there are actually a lot of subtle points that need to be addressed when building the perfect one. First of all, we need to keep in mind that the self-extractor is what the user is going to see, not the (now embedded) installer. Therefore, the ability to customize its icon is quite handy and professional. But there’s more.
Here’s my checklist for capabilities in a self-extracting executable maker, all of which WinRAR excels at:
- Silent mode – this is the ability for the self-extractor to be completely invisible to the user while decompressing an archive. If the archive is small, you don’t need to bother the user with yet another welcome dialog, where to extract to, and/or a progress dialog.
- Extract to temporary directory – hand-in-hand with silent extraction is extracting somewhere that won’t get in the way. WinRAR has the ability to extract to the system TEMP directory, and even cleans up after itself when the installation is complete.
- Ability to execute a file in the extracted archive – this is absolutely essential for building an installer wrapper. Some self-extracting archives are intended just for distributing files and such, not actually installing a program; but for our purposes, we need to have a way to launch the bootstrapper.
- Custom icon – give your installer a professional look by providing your own file icon that can be seen in the web browser downloads and on the desktop.
- Choose whether the executable requests elevation – installation generally requires admin privileges, but you don’t need to authenticate yourself right away. I prefer to defer elevation until the user actually clicks the Install button, which means the extractor doesn’t need to request elevation. If you’re installing from a bootstrapper though, requesting elevation from the get-go could mean only one UAC prompt rather than several.
There are also a ton more options such as the ability to customize the dialogs and logo and display a license agreement–there are seven tabs worth of customizations for self-extracting executables alone!
To put it plainly, WinRAR is well worth the $29 single-user fee. The self-extraction options alone are worth the price. However if even that is too steep for your tastes, there are some free options available, like 7-Zip and IExpress, but if you want the whole package, and my advice, go with WinRAR.
Next: How to write a simple but flexible bootstrapper using the awesome AutoIt scripting language!
Update: I just discovered that WinRAR self-extracting executables even support command-line argument passing! This means I can pass arguments to my WinRAR self-extractor and it will pass them along to the bootstrapper, which (depending on the bootstrapper) can even forward them onto the actual msi, allowing you to define variables that can then be interpreted by the installation engine! Awesome!