No, I meant of course that vurtun + all the callbacks are going to be at least as fast as iod. You'd basically pass a bit of context to every callback, so that

No, I meant of course that vurtun + all the callbacks are going to be at least as fast as iod. You'd basically pass a bit of context to every callback, so that the app would know where exactly it is in its parsing. It's clearly a bit more of leg work in terms of coding, but I would personally take that over introducing a dependency on a larger abstract framework. I come from the embedded programming background, so I would opt for a pre-processor step that takes in a struct definition and generates a proper set of parsing callbacks. This way you'd at least see the interim code before it gets compiled. But I digress, to each his own and that's not a point here. I spent a lot of time optimizing C code and I find your blanket performance claims to be too absolute and cocky for comfort.