3 ms·

An interesting approach: https://github.com/automerge/automerge https://github.com/automerge/automerge Academic project, but from an experienced implementer.

An interesting approach: https://github.com/automerge/automerge https://github.com/automerge/automerge Academic project, but from an experienced implementer. Though there are things missing, I suspect for some applications this might be usable already. The "Caveats" section seems like it may be needlessly scary: > No integrity checking: if a buggy (or malicious) device makes corrupted edits, it can cause the application state on other devices to become corrupted or go out of sync. I wonder if this can be addressed for now at the application level, on top of automerge? > No security: there is currently no encryption, authentication, or access control. This puzzles me: why would a data structure have encryption, authentication, or access control? Perhaps this refers specifically to Automerge.Connection? But I assume any gossip protocol could be substituted fairly straightforwardly? Otherwise I can only presume that by this he's just saying that the TRVE research program is incomplete -- but as far as I know, almost no software meets that sort of high standard! Also, it doesn't mention in that section that currently automerge never truncates data. That's a constraint I believe Kleppmann would like to relax in the future.

thingification 7y ago
> I wonder if this can be addressed for now at the application level, on top of automerge? On reflection, I think the answer to this is very likely "no". I had in mind storing the deltas in a log-structured database -- which I believe is what hypermerge does -- and detecting invalidity at application level and then rolling back to old log records in case of trouble -- but of course that doesn't help with losing sync, or effective loss of data from rolling back or corruption that leaves the data valid, and it's just not very practical. So, even though I suspect the problems in practice are rare (Martin's clearly a very able programmer), I guess that until he gets around to the integrity work he has in mind, it's limited to apps where the data is very transient for now. Those do exist: say, off the top of my head, a tool for collaborating on sketching simple diagrams -- in fact their pixelpusher demo is quite similar to that idea! I think he's currently focusing on performance improvements. Hope to see it grow into that!

thingification 7y ago

> I wonder if this can be addressed for now at the application level, on top of automerge? On reflection, I think the answer to this is very likely "no". I had in mind storing the deltas in a log-structured database -- which I believe is what hypermerge does -- and detecting invalidity at application level and then rolling back to old log records in case of trouble -- but of course that doesn't help with losing sync, or effective loss of data from rolling back or corruption that leaves the data valid, and it's just not very practical. So, even though I suspect the problems in practice are rare (Martin's clearly a very able programmer), I guess that until he gets around to the integrity work he has in mind, it's limited to apps where the data is very transient for now. Those do exist: say, off the top of my head, a tool for collaborating on sketching simple diagrams -- in fact their pixelpusher demo is quite similar to that idea! I think he's currently focusing on performance improvements. Hope to see it grow into that!