I'm currently researching what kind of modelformat i'd like to use. For now i've taken a look at collada and fbx.
While collada is easy to parse (since it's using xml) the data structure is terrible (since it's xml). I didn't want to bother with it's crazy layout so i checked what fbx is like. At first i was impressed. Easy to read and identify the data structure. When i started parsing it, i realized what a crazy amount of bullshit this is. I certainly don't want to bother working hours with coming up with a parser that is able to handle all of fbx's quirks and not error out in one of its inconsistencies.
I want to check a couple other formats before i go back to hustle my way through the fbx data, like md5. Although this requires an implementation of quaternions, which is further delaying the actual goal.
Meh, this is frustrating and annoying. Why are the official model formats in such a bad shape!? Aren't there actual programmers working on it? Don't they know what they do to the community by providing atrocious formats like that?
What's your experiences with modelformats, which do you prefer(ha! as if) or hate?