I happen to be in the middle of writing a simple tar implementation, and had in fact just started playing with the idea of just using it as an on-disk filesystem format. The caveats I can see are that 1. it's approximately append-only[0], and 2. you have to replay the whole thing in order to make sure you have the latest version of a file (I'm leaning towards doing so on mount because then I can also build an in-memory filename-to-block-address cache). Have you hit any issues, or is this as surprisingly reasonable as it sounds?
[0] I mean, if the new version of a file fits in the same number of 512-byte blocks as the old version you could update in place, but that's an unreliable condition at best and also guarantees data corruption if you lose power halfway; really going append-only makes the implementation simple and also makes it easy to be really resilient.
[0] I mean, if the new version of a file fits in the same number of 512-byte blocks as the old version you could update in place, but that's an unreliable condition at best and also guarantees data corruption if you lose power halfway; really going append-only makes the implementation simple and also makes it easy to be really resilient.