Publish to IPNS from Circle CI

[harlantwood]

When the code for this repo changes, we rebuild the JS bundle on Circle CI, check everything into IPFS, publish to IPNS (always the same peer ID / private key), then request all assets from the gateway.

See test below: I changed the text of the exists column header to (exists) and republished:

@jbenet lots of this should be reusable for autodeploys of the website, and whatever else ... ipfs-inactive/website#17 & ipfs-inactive/website#75

[RichardLitt]

Nice!

[jbenet]

This is great! though publishing to IPNS wont work, unless you run CI every few hours. the reason is that IPNS needs to be republished over time.

We'll be changing that to allow other nodes to re-share records-- it's a major protocol issue.

[harlantwood]

OK... Options:

1. Publish to ipns hourly. Maybe by triggering CI, or from a cronjob, eg on heroku or digital ocean
2. Switch to pushing IPFS hash to dns bridge. This requires having a digital ocean token present, so @lgierth doesn't want to have it run from CI. Perhaps it could run from one of our digital ocean servers.

[harlantwood]

@lgierth and @jbenet do you have preferences between options 1 & 2 above?

[ghost]

Publish to ipns hourly. Maybe by triggering CI, or from a cronjob, eg on heroku or digital ocean

mh shouldn't the republisher take care of that? @whyrusleeping

[ghost]

mh shouldn't the republisher take care of that? @whyrusleeping

ooh, the publishing daemon is ephemeral, within circleci. tricky.

[jbenet]

@harlantwood both SGTM. if we do a service we host, then:

• CI could hit a service with the ipfs-path, and it could publish to ipns
• forget CI, and a webhook could hit a service directly who does all the work. (ci is fine with me)

[harlantwood]

I'm going to leave this PR open for now, in hopes that this will change:

@jbenet said in #11 (comment):
publishing to IPNS wont work, unless you run CI every few hours. the reason is that IPNS needs to be republished over time.
We'll be changing that to allow other nodes to re-share records-- it's a major protocol issue.

[RichardLitt]

@lgierth When do we think that will change, for IPNS? Is that in the roadmap?