@falsifian@www.falsifian.org Regarding your last paragraph: Back in December 2020, we already once changed the hashing. I think that was my first contribution, breaking everything by switching to RFC 3339 for the timestamp format. ;-) I’m computing two hashes in my client, the old and current one. And then I just select whatever matching parent exists to build the thread tree.
I could do that again in my client, but you’re right, it’s a different story for jenny. If I’m not mistaken, In-Reply-To could contain several hashes, but the Message-ID header is the issue.
By increasing the hash length for a potential future change, clients could tell, which algorithm to use.
Maybe we could define a magic timestamp in the future that marks the cutoff point. Use the current implementation for messages authored before that magic date or the new algorithm for all messages after that.
But eventually, all clients have to be updated. There’s no way around that, I believe. Simplicity is key and my magic time already adds complexity. :-/