Moving to a state machine-based parser for MARQUEE may seem like an odd decision for an HTML renderer. You would naturally think and perhaps common sense would dictate that regexps are enough for something simple (they aren't) or that you absolutely need a tokenizer + recursive descent parser (factually not true).
Ideology, tradition, and beliefs apart, the main reason to choose a state machine-based parser is this: A state machine can be implemented on any architecture. And on small machines and virtual machines, you get absolute control over every character and every resource used, nothing goes to waste.
This approach works the same on a powerful server, a fancy latop, or the same code but targeting a microcontroller.
Or an old 8-bit machine.
