From @githubber314159 on Sun Apr 16 2017 18:43:46 GMT+0000 (UTC)
What is the best way to run dynamic applications?
Can we just take the raw files from a DBMS and publish a snapshot each time a COMMIT; is issued?
Or should DMBSs accomodate to the new p2p-way the data is stored in deeper layers?
Is it better to have a central database from which authoritative snapshots are distributed amongst the peers?
Or is it better to have distributed databases running independently from IPFS altogether, resulting in a complete separate system? Databases are powering the web of today, not static files. We need to deal with the database issue. See semantic web for related discussion.
What are the most pressing issues to come up with a solution for this?
Are we at the stage of incomplete designs of our ideas?
Are we at the stage of incomplete implementations of our designs?
Are we at the stage of incomplete advertisement of our solution?
Please let me know, I’m interested what your answers will be.
I think this is a really important question, it might be worth also asking what apps we see running decentralized and what centralized.
For example … full relational DBMS on large files might be very hard to run decentralized BUT smaller object-store applications, where things are stored by their id might work fine in the decentralized web.
However it needs the mutable/immutable layer to work well, which is a huge limitation in IPFS (since its only working on some platforms, so cant really be used in applications).
In my own design thinking, the other thing that drives the web is lists … for example the items in a news feed are a list, as are the Likes on each item, in some people’s thinking I’ve heard these called “Append Only Logs”. I believe a significant set of applications map easily onto that kind of structure. It can be hacked in IPFS (I’ve been doing it via IIIF) but its a real hack.