Try this:
Open a command prompt window
Type "netsh" and press enter
Type "help" and press enter
Play around a little bit; jump around to different contexts, check syntax for a couple of commands
Appreciate the amount of power netsh gives you.
Appreciate how much time I've spent translating all them syntax messages to Swedish.

As a user, I think it's great that we keep making it easier to use Windows from the command prompt. As a localizer, I'm not as thrilled... localizing command line utilities isn't all that straight forward.

One thing to note is that these messages are really two-dimensional objects stored in one single string. This is how one single localizable resource is stored:

\nUsage: %1!s! [name=]\n             [[routing=](enabled|disabled|default)]\n\nParameters:\n\n       Tag              Value\n       name           - Interface name.\n       routing        - Whether to act as a router.\n\nRemarks: Sets 6to4 interface configuration information.\n\nExamples:\n\n       %1!s! "Private" enabled\n\n\n

And this is what it really represents:

Usage: %1!s! [name=]
             [[routing=](enabled|disabled|default)]

Parameters:

       Tag              Value
       name           - Interface name.
       routing        - Whether to act as a router.

Remarks: Sets 6to4 interface configuration information.

Examples:

       %1!s! "Private" enabled

(This is one of the shorter examples. Some are several thousand characters long.)

The markup in the string doesn't cause me much problems - I have an editor that can show the string as it's mean to be seen. Still, there are a good few things I need to keep in mind.

1) Row length. One of the constraints we place on ourselves is that any row in a syntax message should not be wider than 80 characters (we aim for 78). Since the default width of cmd.exe is 80 chars, if a string contains a row that's too long, the text will be uglily wrapped. Because of this constraint, we may need to insert extra rows instead. It's typically easy to make sure to follow this rule, and it's fairly straight forward to write a script that scans my translations and reports any violations. Except of course when the string contains a placeholder. There are syntax messages with examples that will expand beyond 80 chars in runtime. It's therefore pretty handy that the whole message is one single resource, as I can add and remove rows as needed.

2) Alignment. Above you have a nice table with two columns. We need to keep it looking nice. This isn't always easy. If the first column expands in width, the second (which usually contains more text) needs to get narrower and this can look ugly too. Also, one needs to make sure not to mix spaces and tabs when aligning as the end result might surprise you. I don't have a tool that can spot misalignments, but it's been (way down) on my todo list for a while.

3) What do I really translate? Looking at the first two rows, only the words "Usage" and "string" should be translated. Anything else needs to be left untouched since the actual options are hard coded. Over translation here doesn't have functional implications, but it can confuse the user. Now, this is a pretty straight forward example, but there are cases when it's not at all clear what parts I shouldn't touch.

4) Recycling & consistency. These messages aren't recycling-friendly. Since aligning strings here means breaking up sentences over several rows, it's very hard to recycle a part the string. You often end up re-translating the same thing over and over again, and this will give inconsistent translation. Not the end of the world, but it sure is a tedious way to waste time. Also on consistency, make sure that the items in the Tag column match whatever comes after "Usage" at the start of the string and the examples at the end.

5) Spell checking. Spell checking this junk is hard -- a lot of elements are left in English, sentences are broken up over rows... not fun.

We struggle with all of these issues for every release. I've gotta try and do something about that...


Btw, one thing that's good with these tools is that they can be automated. So, it's entirely possible to write a script to call all kinds of command line utilities, analyze the output and find problems.