Moreover, if I were typing a list containing (a), (b), (c) and ended up with a copyright symbol, I’d probably end up wasting a lot of time hunting through my application preferences looking for some obscure autoreplace/autocorrect feature I’d missed rather than thinking to turn off calt or dlig.
André, isn’t the <frac> feature quite straightforward?
The method you outline is certainly commonplace, but when applied at the paragraph level it entails some guesswork on the part of the application (i.e. excluding things like 2018/01/07) and so forth. That guesswork gets considerably more complex if the font allows the creation of fractions from things like x/y or 2n!/z, and switching from a font which supports only numeric fractions to one that supports additional fractions may well result in text becoming mangled. Also, until applications start supporting type 8 lookups (reverse chaining) Tai Leming’s approach involves imposing an arbitrary length restriction on fractions.
This is why I’ve always preferred fonts which require you to select the thing you want to be a fraction, not the paragraph as a whole. I suspect I’m probably the minority in this regard, but I don’t want my fonts trying to guess my intentions (just like I don't want MS Word trying to guess which kind of quotes I want).
I’ve always wished there were some single, global preference “leave my text the $%& alone” which all applications would obey, and which would turn off absolutely everything related to smart-quotes, auto-correction, auto-formatting, auto-anything-else. In other words, just have the system assume that what I typed was what I intended to type.
If people start trying to implement these sorts of things in GSUB tables, that creates yet one more place where I have to turn things off.
Even worse, it creates a situation where the behaviour of your text may be entirely different depending on which font you choose to use. This is already a (relatively minor) problem in that different fonts often implement various features differently — some examples that come to mind are 'frac' and 'ordn' where some fonts allow you to apply these features to runs of text and then try to figure out what you intended contextually, whereas others expect you to apply the feature only to the characters you intend to be affected.
Such inconsistent behaviours are undesirable from a users point of view and I think that inconsistency outweighs any convenience which might be achieved by having the font try to guess your intentions.
U+02BC is named ‘apostrophe’, but this character is classified as a modifier letter rather than as punctuation. I think this was intended more for words like ’alif or O’odham where it represents a glottal stop. (Similarly U+02BC should be reserved for words like Hawai‘i).