> My reaction after reading this is that if it's too much burden for the original author to keep up with updates, it's very hard to make a case for people to self-host their own services using Sandstorm. Upgrade effort will now get multiplied many-fold as everybody has to duplicate the effort that one maintainer was putting in.
That's not correct.
Self-hosted sandstorm is completely auto-updating. You don't have to do anything. I still have to push updates periodically, but I intend to keep doing that. After I push an update, all Sandstorm servers update themselves automatically within 24 hours.
Updating Oasis was, ironically, much more work for me than pushing an update to self-hosted sandstorm. Oasis is built on a much more complicated cluster architecture designed to be scalable -- scale that, sadly, we never actually needed.
Kenton, pardon my hyperactivity, but here is a couple of observations you may find useful:
>Oasis is built on a much more complicated cluster architecture designed to be scalable
You could dog food Oasis on top of Sandstorm. For example, Oasis could be a Sandstorm app and use Sandstorm services. In this way, you could feel very closely where and when Sandstorm is lacking. And once you fix or invent those parts, the results would be available to all of your customers, and not exclusively to Oasis. This would shift the laborious spot from Oasis to Sandstorm, and would allow you to open up the enterprise gates.
The second observation is a model of sales. You sold a "decentralization idealism" as you coined it, but with an optional central place like Oasis. In my opinion, this offering represents a natural conflict and is not destined to work in a sustainable way. Why not sell a good old license for advanced/enterprisey features? Oasis would be a great add-on to that model for those who wanted to rent the resources. And yes, exclude the free tier from Oasis to make things fair.
Oasis is running Blackrock, which is basically a cluster architecture version of Sandstorm. And while it was originally an enterprise/paid offering only, Blackrock is also now open source. However, Blackrock was really never built out as far as intended, and really only runs on Google Cloud.
There's a fair bit of history you may not be aware of. For instance, the payment system used to be part of Blackrock, but then was moved over to Sandstorm when it was open sourced so you didn't need to run Blackrock to sell subscriptions to a Sandstorm server. And there used to be paid-only features, but when Sandstorm-the-company ran out of money, they were made free as well.
That's not gonna work. Sandstorm is explicitly not designed to be big-cloud infrastructure, it's designed to host small-scale apps. Oasis is intended to be a big Sandstorm server that utilizes a cluster of machines. It therefore needs to sit on top of big-cloud infrastructure, not other Sandstorm servers.
> You sold a "decentralization idealism" as you coined it
That was teleclimber's phrase, not mine.
> In my opinion, this offering represents a natural conflict and is not destined to work in a sustainable way.
I agree, if I were doing it over I don't think I'd build Oasis. I'd focus instead on federation and migration features, and partnerships with hosting providers and device manufacturers to make setting up a private Sandstorm server as trivial as possible.
> Why not sell a good old license for advanced/enterprisey features?
That's not correct.
Self-hosted sandstorm is completely auto-updating. You don't have to do anything. I still have to push updates periodically, but I intend to keep doing that. After I push an update, all Sandstorm servers update themselves automatically within 24 hours.
Updating Oasis was, ironically, much more work for me than pushing an update to self-hosted sandstorm. Oasis is built on a much more complicated cluster architecture designed to be scalable -- scale that, sadly, we never actually needed.