For one thing, entry-level devices sold by Sprint MVNOs are less likely to support 4G, and devices that don't support 4G still store the subscriber identity on the device. For another, I was under the impression that CDMA2000+LTE devices sold by Verizon, Sprint, and Sprint MVNOs were carrier locked not to accept a SIM from a GSM+UMTS+LTE carrier. And finally, can you swap CDMA2000 subscriber identities by swapping SIMs in a CDMA2000+LTE phone, or is only the LTE subscriber identity stored on the SIM and the CDMA2000 subscriber identity on the device itself?

Swap them. Do they still work?

Such as a count of each distinct item on an order and a count of cents that the customer paid for each item. For obvious reasons, a negative value would be inconsistent.

Ordinal numbers have two mappings to the cardinal numbers. A "zero-based count" represents an ordinal as the number of items that have already been considered: zero, one, two. A "one-based count" represents an ordinal as the number of items that will have been considered once processing of this item is finished: one, two, three.

True, I can find another language. But the other language just happens not to have a runtime environment on the target platform, as 0xdeadbeef and I pointed out.

The point remains the same nevertheless. MIDlets for Java ME phones have to be written in pure Java for the same reason as applets for browsers supporting Java SE.

I don't see a single way Windows differs from Android in this respect. You can't plug a SIM into a Windows PC with a Sprint mobile broadband adapter for the same reason that you can't plug a SIM into an Android phone sold by Sprint. A device that takes a SIM won't work on Verizon or Sprint or a Sprint MVNO either because the U.S. CDMA2000 carriers have chosen to program the subscriber identity directly into the mobile equipment rather than using a CSIM. This is true no matter whether a device runs Windows, OS X, desktop Linux, Windows RT, iOS, Android, BlackBerry, GreenBerry, Three-a-MeeGoes, or whatever.

Except Snap. Windows and Windows RT both offer the ability to split the screen down the middle and run a web browser on one side and a note-taking application or word processor or whatever on the other. Windows has had such a tiling window manager since version 1.0 in 1985, back when the original NES was hot $#!+. The Mac had floating mini-apps (called "desk accessories") since its launch in 1984, long before even MultiFinder. Android, on the other hand, runs a window management policy of all maximized all the time. My Nexus 7 tablet's screen is bigger than two Android phone screens put together, but it can't run two applications each in a phone-sized window because the Android CDD allows applications to assume that the screen's size will never change after installation. Why must a calculator app fill your laptop's entire screen and cover up the document containing the numbers that you're adding?

What you say is true of big games from big companies, not so much of small games from developers that haven't yet found a console publisher. Those are more likely to be released on OS X, desktop Linux, Android, or iOS before one of the major consoles.

So does the USB mobile broadband dongle from any carrier that isn't Verizon or Sprint or a Sprint MVNO. The problem here is that one has to pay hundreds of dollars a year to use that SIM card unless applications are designed for "offline first", anticipating that a user's Internet connection will be intermittent. I imagine that Windows applications are more likely to support offline use than Android applications, especially because Android was first sold on phones.

Laptops have had a front-facing camera as a standard feature for years, and they've had either VGA out or HDMI out or both since I started using laptops in 1999.

No it isn't. The Android 4.3 update broke the ZAGGkeys Flex and several other Bluetooth keyboards that worked under Android 4.2, mistakenly recognizing them as "non-alphanumeric keyboards" (that is, gamepads). And it can't be fixed without wiping and rooting the device to rename a keyboard layout file.

Provided that the platform curator even allows the use of lower-level languages. For example, Java applets have to be written in Java, and Xbox Live Indie Games and Windows Phone 7 applications have to be written in C#.* An applet that attempts to use JNI or an XNA game for Xbox 360 or application for Windows Phone 7 that attempts to use P/Invoke will die with a security exception.

* Technically, XBLIG and WP7 allow the subset of verifiably type-safe CIL accepted by the .NET Compact Framework. But in practice, languages other than C# either aren't verifiably type-safe (such as standard C++ in C++/CLI) or require library facilities not present in the .NET Compact Framework (such as any DLR language).

If your application logic's requirements include being able to represent values between 2^((2^n) - 1) and (2^(2^n)) - 1, such as 128 through 255 or 32768 through 65535 or about 2.1 to 4.2 billion, in a cache-efficient array, you usually want to use an unsigned type. This often comes up when trying to represent the native unsigned data types of an emulated machine or the unsigned data types of various SQL databases. You could use a type twice as wide, but that'd fill L2 cache twice as fast, causing capacity misses. And on mobile, it'd fill RAM twice as fast, causing the system to kill your application for having run out of memory.

By having the object code format be JVM bytecode, CIL, or LLVM bitcode.

At that point, why not just code for the Nintendo Entertainment System? It has about the same amount of RAM as an Arduino Uno, and your potential employer's HR department likely grew up playing it. In any case, either this or the NES is likely a useful counterpart to certain Slashdot users who claim developers of games for these limited platforms are "living in the past".

Apart from memory connected to the video controller, the Nintendo Entertainment System has 2048 bytes of RAM. (Many games, especially later ones, have an extra 8192 bytes on the Game Pak PCB to store large destructible maps.) The ATmega328 in the Arduino Uno also has 2048 bytes of RAM. The ATmega2560 has 8192 bytes, like a Sega Master System. (Source) Tricks to use memory more efficiently include byte-sized variables and even bitfield variables.

With the entry barriers and platform fragmentation that exist in several areas of the software market, I imagine that a lot of smaller shops not big enough to be able to afford to hire specialists in multiple platforms will have to disappoint users of at least one major platform with a notice "We are seeking a publisher to bring (name of product) to (name of platform)."

The trademark issue that you mention is orthogonal to the laws of war issue. Trademarks must be respected in all games, be they realistic or fantasic. Laws of war (other than the trademark-like laws protecting the Red Cross logo) need be respected only in realistic games.