add support for 'real'-s3 and redis

[gretel]

so one could have the benefits of rebund without keyfile :smiling_imp:

[mezis]

Hi @gretel, I'll probably leave this issue open and/or mention in the README :)

It's a chicken-and-egg problem: the point of rebund is to not have to download and build a bunch of gems (or other software) on each build.

The problem with S3 (or Redis) is that interacting with them typically requires some support software. In the case of S3, doing so in Ruby requires quite a few packages (that would need to be installed before downloading, or checking, an existing bundler tarball); installing them before calling rebund would largely defeat its purpose (you'd spend a minute or two installing aws-s3 for instance).

This said:

• It's possible to use S3 directly from a Bash script though, I haven't explored that option yet.
• In the case of Redis, assuming redis-cli is available, I suppose it would be possible as well to get payloads just using Bash and the Redis wire protocol. I'm not sure using Redis for this is very effective though, as the cost of memory is high (i.e. I couldn't run this for free).

It's also pretty straightforward to deploy and use your own Keyfile server...

Any thoughts for next steps?

[gretel]

hej @mezis, now i get it, thanks for explaining. :+1: for adding your comment to README. for me it is the other way around. s3 and redis are very cheap while setting up keyfile would bring up more administration hassle (really, nothing to say against keyfile :love_hotel:) than i am willing to take in this case. so i would trade cost of redis for less (human) time spent waiting for some CI processs. next step: try to use yarp in production only, should fail if index older than Gemfile locked is.

[mezis]

By the way — if you want a key to use the default Keyfile app (which I manage at http://keyfile-production.herokuapp.com), just ask.

[gretel]

well alright let my try please :+1:

[mezis]

Just emailed you credentials.