I was playing around with PostScript encoding vectors to make a Thai
hack Unicode font (Lannaworld) usable with GhostScript when I noticed
that the Postscript names for the glyphs of sara a and mai han-akat
(short /a/ in 'open' and closed syllables respectively) were out of order.

These glyph names (more precisely numbers) are effectively registered
under ISO/IEC 10036, which was handled by the Association for Font
Information Interchange (AFII).

For sara a, U+0E30, the glyph name was afii59729.
For mai han-akat, U+0E30, the glyph name was afii59728.

Apart from these two, the glyph number differs from the TIS-620 and
Unicode codes by a fixed offset. (This implies that glyph numbers
have been allocated for the representative glyphs of the as yet
unallocated U+0E3B to U+0E3E!) The same reversal is visible in the
Tahoma and AngsanaUPC fonts, while there is no such reversal in the
Titus Cyberbit Basic or the registry page at
http://10036ra.org/glyph-table.html?page=597 .

Does anyone know who is error, and which scheme one should adopt for
fewest problems?

Richard.