I agree with the main thrust of the article, that we should instead trust the transport mechanism rather than the cryptographic properties of SHA1, but because of this I don’t really think switching the default will be a costly mistake. I rarely use full SHA1 hashes right now; I only use the truncated version and I don’t think any user cares about the length of the full hash. As for compatibility with forges, it’s just a small UX problem that should be solvable: don’t let the user choose the format when a repo is created; instead choose it when the first push happens.