you're saying and asking too many things at once that i don't fully understand yet.
the script is a work in progress and definitely not perfect but i think the rat is not a good
example of the issue because it is too simple, you must look to the character models,
like camoufla-dc_dododo etc
i don't know what you mean here:
creating groups for the rat all modulo 100 fi numbers created none overlapping uvs
"all modulo 100 fi numbers" -guess i have to look that up.
i think the *.tri file might be the key to material groups here, for the rat the UVs produced by the
script in Noesis look like 2 overlayed sets and 00713a4f.tri has two 160 byte blocks of data.
When I export the rat with your script from Noesis to obj the vns are bogus
likely yes because like i said i wasn't sure how they are handled here,
i could just avoid the normals but i was just playing around trying to get them working,
you should probably just comment that part out of the script (line 44 at least).
i don't know what to do about the winding issue, maybe they are really quads?
in the link you gave the post as of Thu Jun 09, 2016 12:10 pm is the one you referred to?
just click the link and it will take you directly to slowerthanlight's post,
the info i was reading is near the bottom of that post.
basically decode 2 shorts into 3 floats and run some algo on them,
i thought decodeNormals32() probably is what this was for.
rapi.decodeNormals32() is briefly described in the __NPReadMe.txt
I made use of an existing C routine and it seems to work
can you post this routine?