In the MEL help file, the "OEM Translation" topic says that, in Tools | Customize | Editing, the OEM /Ansi choice controls not only the font type but also the behavior of Alt-Numpad, e.g. how Alt-164 is interpreted. It says this is necessary because Windows translates Alt-Numpad codes from OEM to ANSI. It also says that, in OEM mode, MEL applies a reverse translation to undo the Windows translation.
I found that MEL is doing this as described. However, there are two problems with the reverse translation feature: (1) The map between OEM and ANSI is not fully intertible and (2) Windows has a feature, apparently unknown to MEL developers, that makes this feature unnecessary. If you add a zero before the numpad code, e.g. Alt-0164, Windows does not translate the numpad code – it gives you the actual code you entered.
I suggest that MEL drop this reverse translation in OEM mode and instead rely on the Alt-zero-code feature in Windows.
This same help topic also references an offline 1995 article on OEM, ANSI, code pages, and Unicode. There are numerous online sources on this topic, such as the Wikipedia article on code pages at http://en.wikipedia.org/wiki/Code_page .
The help topic describes Unicode as a two-byte character set, which has not been true since Unicode version 2.0, released in 1996. Unicode is a four-byte character set, with over a million possible code points. Only 64K was possible in the two byte scheme. The four byte scheme can be encoded in two bytes, and usually is, using the surrogate character scheme introduced in Unicode 2.0.