A great book tangential to this subject matter is Jaron Lanier's "you are not a gadget" which spends a substantial amount of time on the development of music in parallel to computers. My recollection is that midi basically became the dominant format because it was first, and it was first because it was more or less the most obvious way to represent music in a simple programmable format. Other formats that allow for more flexibility didn't crack through due to the critical mass of adoption of midi, so it persisted until today.
This is not really correct.
The objections to MIDI lie in its simplification of musical pitch (which isn't 100% baked into the specification, but does turn out to be overwhelmingly common in actual practice).
This article is mostly about timing, which MIDI doesn't have a huge amount to do with, and the specification doesn't really have any particular problems in this area (it's a series of timestamped events; the resolution of the timestamps is arbitrary, so MIDI can precisely represent incredibly wonky or incredibly precise musical timing).
There are also issues with MIDI 1.0 if you want to provide more performance information on a per-note basis, as happens when playing many instruments. There is an excellent workaround for this with MIDI 1.0 (called MPE - MIDI Polyphonic Expression) but people wanted more and these days MIDI 2.0 mostly provides it.