Allow mark glyphs and mkmks to form a tree structure

[davelab6]

@be5invis should https://twitter.com/belleveinvis/status/1071354747863556097?s=19 be a proposal, rather than an ambiguity? :)

[tiroj]

Would be helpful for a proposal or further discussion of this issue to include a non-programmer explanation of the question. I sense the term 'tree structure' has a meaning here that I'm perhaps not understanding. @be5invis' tweet mentioned that Andrew Glass' Egyptioan hieroglyphic font looks like it has been built using a tree structure—and I have a good understanding of how that font works and fonts I have made using similar techniques—, but from the subsequent twitter discussion with @behdad it seemed to me that there's perhaps some technical meaning of tree structure as applied on the shaping engine implementation of GPOS that I am not understanding.

[be5invis]

It's been a while :) The idea is that each glyph could define a mark anchor and various base anchors. When placing a new glyph, it will attach to the nearest glyph with the base anchor that has the same anchor class with the incoming glyph's mark anchor, and is not occupied. If such glyph could be found then a mark placement is performed. With such placement mechanism a root-first sequence of glyphs could form a mark tree.