Martin Kleppmann

You describe parents as "a set of operations that this transaction depends on" — presumably you mean a set of *transactions* that it depends on? We also need to specify...

> The parents field lists the indexes of previous transactions. This is simple, straightforward and I think it should work for any format. Ah yes, that's good. > Perhaps a...

More substantive comments later, but for now: great stuff, and bravo for the excellent term "Rumpelstiltskin rule"!

The reason the sender suppresses sending a change it's already sent is that sometimes the recipient asks for the same change several times in a row, and we want to...

I believe this is happening due to [this condition](https://github.com/automerge/automerge-rs/blob/0ab6a770d82785464043505a548a6f41cb593e0b/rust/automerge/src/sync.rs#L30-L34), which suppresses the sending of a Bloom filter if there are currently missing dependencies that are not one of the remote...

An API to translate between a cursor position as character ID and an index would be a good thing to add. It would first need to be added to the...

There is a project in progress (#706) to add a move operation that will allow array elements to be reordered. For now, a workaround you can use is to give...

Hi @kirs, what's your use case for wanting a bounded counter? I think it would be tricky to implement in Automerge, because the process of transferring a quota from one...

Love this! Some thoughts: - Should we allow a block to be an arbitrary map object? I feel the proposed structure (with `type`, `parents`, and `attrs` fields) is perhaps a...

I would be open to changing the default behaviour so that every assignment generates an op, regardless of whether the value has changed. I doubt that many applications are relying...