Thanks for the link.

我完全同意你 quote 的那句話, 也在文中補上了 UTF-16 較 UTF-32 節省空間.

但我還是認為用 UTF-16 不如採用 UTF-32. 若真要節省空間, 尤其是在大部份字元是在 ascii 定義範圍之內, 如程式碼, 那更應採用 UTF-8.

程式好寫用 UTF-32, 若空間有限則選 UTF-8.

[Update: Mar 3, 2009]

這句話一點都沒錯:

It’s wasteful to use 4-byte wchar_t strings to carry UTF-16 data, …

問題是為什麼要存成 UTF-16? 用 UTF-32 (wchar_t) 或 UTF-8 (char) 就好了啊~

// On many other platforms, sizeof(wchar_t) is 4 bytes by default. We can make
// it 2 bytes by using the GCC flag -fshort-wchar. But then std::wstring fails
// at run time, because it calls some functions (like wcslen) that come from
// the system’s native C library — which was built with a 4-byte wchar_t!
// It’s wasteful to use 4-byte wchar_t strings to carry UTF-16 data, and it’s
// entirely improper on those systems where the encoding of wchar_t is defined
// as UTF-32.
//

其中，”It’s wasteful to use 4-byte wchar_t strings to carry UTF-16 data, and it’s entirely improper on those systems where the encoding of wchar_t is defined as UTF-32.”。

看起來似乎與站長的見解不盡相同。註：本人對於UNICODE沒有研究，只會呆呆地用std::wstring and wchar_t 來寫UNICODE 程式，^^!

Here is a link to that string16.h

http://src.chromium.org/viewvc/chrome/trunk/src/base/string16.h?revision=1464&view=markup

第二個回應的歷史我大致清楚.

針對第一個回應我還有接續的問題, 寫在 update 裏面. 感謝解惑.

VC6不夠新，所以wchar_t不是build-in/primitive type，也因此用ostream時，comiler無法分辨wchar_t與short，而無法印出wchar_t字元。

也因為VC6的wchar_t不是build-in/primitive type，所以後來的MS C/C++ compilers，預設wchar_t也都不是build-in/primitive type。不過，可以在project setting裡調整，將wchar_t試作build-in/primitive type就是了。