(My solution can more or less be summarized as the 'ACL' consisting of a list of mutable DHT pointers in a single-owner data object, where clients retrieve-and-overlay the current object at each mutable pointer, and the owner eventually reconciles them centrally when online.)
-
-
(So non-owner writers can independently publish changes to the original data object, and every once in a while the owner combines all of the non-owner changes into the canonical copy, invalidating the overlays, so as to limit client overhead.)
1 reply 0 retweets 0 likes -
(The main issue with my approach - and I'm really using too many parentheses - is that it's easy for one of the overlays to get lost due to insufficient seeding, thereby resulting in an incomplete changeset. Wondering if Dat handles this better.)
1 reply 0 retweets 0 likes -
Replying to @joepie91
I haven't looked into that code yet byt, when I was planning on rolling my own in the same vein, I was going to handle authorisation by signing the public key into the DAG as a special authorization operation node. Deauthorisation would similarly have a special node…
1 reply 0 retweets 0 likes -
… I'm assuming multi-writer Dat is using a similar scheme but I'll find out exactly how they're doing it.
1 reply 0 retweets 0 likes -
Right, it's as I'd said (& documented in the HyperDB architecture document):https://github.com/mafintosh/hyperdb/blob/master/ARCHITECTURE.md#authorization …
1 reply 0 retweets 1 like -
Replying to @aral
Ah, thanks, that's a good read. Seems a bit similar to my design, but without the 'reconcile' step - how does it ensure that all the entries remain perpetually available? It seems to me like a lost entry could result in an entire branch becoming inaccessible.
1 reply 0 retweets 0 likes -
Also, it seems like it requires on a protocol level that the full history of every hypercore is kept around in perpetuity? Or am I misunderstanding it?
1 reply 0 retweets 0 likes -
Replying to @joepie91
As far as I understand that is a feature/limitation of the design. You should be able to tombstone content but I don't see how you could ensure authorisation without the full DAG. The protocol itself supports sparse data and metadata. I'm still very new to Dat myself, btw.
1 reply 0 retweets 0 likes -
Replying to @aral
I see, thanks for the commentary :) I'll have a closer look into it some time soon, in particular to see whether the availability properties are better than those of the overlay-and-reconcile model...
1 reply 0 retweets 1 like
Look forward to seeing what you're working on too btw :)
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.