search

ipfs / notes

IPFS Collaborative Notebook for Research
MIT License
402 stars 31 forks source link

Paradigm / best practices for publishing beyond the local node #332

Open benjaminbollen opened 8 years ago

benjaminbollen commented 8 years ago

One recurring question I've had - and the answer of which of course depends on the use-case - but perhaps I might learn something from your responses.

is some api provided to ensure redundancy in storing objects on the DHT? For example, in a controlled environment, can I ensure that some objects are redundantly stored (or subDAGs prioritised by some nodes) after a local node has locally stored, pinned and published them?

yalamerde commented 8 years ago

You mean to work around unpinned, garbage collected hashes at remote nodes?

almereyda commented 8 years ago

Or an ability to pin ipns names?

benjaminbollen commented 8 years ago

@yalamerde and @almereyda (any coincidence the names are an anagram?) thanks for good links to useful preceding questions! yes, largely my question comes down to such questions; in either case it looks to me that a solution is on a higher level of responsibility.

To complement both of your questions: is there a communication channel to pin on remote nodes (whether or not the remote node accepts is a separate question).

jbshirk commented 8 years ago

I'm certainly interested in this topic. I note that since October, my uploads to IPFS have taken at least a day (the last one took 2 days) to get "pulled" deep enough into the network to be reachable by the gateway. I have wished that there were a twitter account for cognoscenti on IPFS to post (non-gateway) links, where others could access them to speed up propagation.

jbenet commented 8 years ago

(took 2 days) to get "pulled" deep enough into the network to be reachable by the gateway

i think that's just (a) NAT traversal problem, or (b) root-is-provided-last problem. (a) can be worked-around by just trying to connect to all bootstrap nodes in the list-- one of them is likely the gateway your dns resolves to. (b) can be worked-around by republishing the root node: ipfs object get $hash | ipfs object put, cutting the provider line. (both of these should work immediately, and yes both will be fixed)

jbshirk commented 8 years ago

@jbenet got a link for bootstrap nodes and where to put them?

jbenet commented 8 years ago

ipfs bootstrap | ipfs swarm connect

(not all will work actually, because some have transitioned to dev0.4.0 -- so may need to manually prune those out before feeding them to swarm connect)