[mrkurt]

It's an attempt to reduce friction. Developers try services like ours out for fun before they use them in anger. Putting a price on the fun stuff keeps people from trying it. It's not that $6.88/mo is too much, it's that >$0/mo is too much for developers who are used to their side projects being subsidized by monopolies or VCs.

People spend a lot of money on infrastructure when they use it in anger. Free apps cost us ~$0.50/mo. If we can sift through 1,000 free users and find one who converts to $25k per month, we're delighted.

[mickael-kerjean]

> If we can sift through 1,000 free users and find one who converts to $25k per month, we're delighted.

Does this ever happen? I was contacted by a F500 who needed some custom version of my open source project (https://github.com/mickael-kerjean/filestash). 25k sounds unreal but hey I have no experience in purchasing enterprise grade software and it's not simple to find actual figures so I was aiming at making an offer at 2.4k/month. Is it a rock bottom figure in enterprise grade software?

[mrkurt]

Ok this is a good and difficult question.

For _our particular service_ this happens, but that's partially because the service is designed to scale to very large customers and the market has set a price for cloud computing resources. Companies already spend money on infrastructure, we're not having to convince them to do that from scratch. We do have to convince them to point that firehose our direction.

What you're doing is more difficult. In general, big companies will spend $10k per month as easily as they'll spend $2.4k per month. It's equally difficult to convince them to do either.

Does filestash get used by multiple people in an organization? If I were trying to price this, I'd invent a "per user" price and target $500/mo for small organizations (like $50/user). Enterprises don't love per user prices that scale, so it gives you a story for "we'll give you unlimited users for $10k per month".

I am more than happy to talk more about this if you want to send me an email, but I'm mostly guessing about OSS licensing. I haven't done the work for real.

[nextaccountic]

Hey, talking about friction.. in your pricing page you say https://fly.io/docs/about/pricing/#managed-ssl-certificates

> Managed SSL certificates

> We use Lets Encrypt to issue certificates, and donate half of our SSL fees to them at the end of each calendar year.

> Single hostname certificates > Free for the first 10 > $0.10/mo for additional certificates

> Wildcard certificates: $2/mo

So, if my pet project needs a wildcard certificate from the get go, I go from $0/month to $2/month before writing a single line of code... can I at least run acme on my own to cut this expense, or this is prohibited?

Actually, to think about it.. Let's Encrypt is free. Is charging for it a significant revenue stream? I would love to see the rationale for that, because it seems to me that this hits small customers disproportionately.

[mrkurt]

The infrastructure to manage and distribute certificates is pretty complicated, and doing support for certificate issues is expensive. This is why we bill for them. More details here: https://news.ycombinator.com/item?id=22624262

For people who need a whole bunch of certificates, the price hasn't been an issue.

The problem with services that are free at any scale is that there's never an incentive to make them any good. Most of our certificate infrastructure improvements have been done because people who were trying to give us money were having a hard time of it.

[nextaccountic]

Fair enough. Indeed I didn't consider support costs. Thank you for your answer!

Actually let me ask another thing. Your FAQ mentions you're considering hosting CockroachDB as a drop-in distributed replacement for PostgreSQL [0], and also you currently offer a distributed, eventually consistent PostgreSQL replication solution [1].

Is either Tikv [2] (distributed key-value store) or Tidb [3] (distributed database with a mysql interface, built on top of Tikv) on your radar?

You already offer Redis as a key-value store, but Tikv has an amazing property: it ensures strong consistency globally (not eventual consistency). Tidb, being built on top of Tikv, also has strong consistency.

[0] https://fly.io/blog/fly-answers-questions/#q-what-is-fly-doi...

[1] https://fly.io/blog/globally-distributed-postgres/

[2] https://github.com/tikv/tikv

[3] https://github.com/pingcap/tidb

[js4ever]

How is it possible that free apps cost you $0.50/mo, the ipv4 alone worth more than that

[mrkurt]

I forgot about the IP addresses, those are not cheap. But we own a bunch and can optimize that cost away really easily too. :)