Thai layout broken for <0E25, 0E37, 0E4C>?

According to http://www.nectec.or.th/it-standards/keyboard_layout/thai-key.htm and http://www.inet.co.th/cyberclub/trin/thairef/wtt2/char-class.pdf, the ICU implementation is correct. They say that U+0E37 has class AV3 and that U+0E4C has class AD1, and that AV1 does not compose with AV3.

Sorry, make that "*AD1* does not compose with AV3." I'm checking w/ IBM Thailand to confirm whether or not the sequence is legal.

Suwit Srivilairith, my contact at IBM Thailand, confirms that the sequence is illegal according to WTT 2.0, which is a Thai national standard. Other posters on the Unicode list thread point out, however, that the strict checking of WTT 2.0 makes it impossible to write Pali and Sanskrit using the Thai script. Another poster suggested that the strict checking of WTT 2.0 should more correctly be done in a spell checker rather than in the rendering engine, which makes sense to me.

Here's the Microsoft Thai OT spec. section about invalid combining marks: http://www.microsoft.com/typography/otfntdev/thaiot/shaping.aspx

Therefore, I'm inclined to think that more relaxed checking is in order. This, unfortunately, will require a redesign of the Thai engine. Perhaps this should be done in conjunction w/ the work for Thai and Lao OpenType processing, and perhaps the generic mark checking. (Tickets 2382, 2386, 4740)