What I Would Do If I Ran Tarsnap

How much of this could be done by building a new business that internally uses tarsnap? The business provides these service plans and completely abstracts the client away from the fact that it uses tarsnap under the hood. The primary risks I can see:

1. Colin gets hit by a bus and Tarsnap disappears. There's not much that this new company could do to protect itself from this problem, short of convincing Colin to implement a failsafe that, upon his verified death, hands full control of Tarsnap (including any relevant private keys and passwords) over to someone who can continue the service (and that someone could easily be the founder of this new company).

2. Technical support that this company could provide would be limited based on the level of support that Colin would provide to this company. Again, perhaps this company could draw up an agreement with Colin for priority support, paying some large amount of money per month for the privilege.

3. There's probably other issues I'm missing, perhaps this company can't make the same guarantees tarsnap itself could (e.g. ability to be HIPAA-compliant) due to not being in control of the technical aspects.

But otherwise, it could certainly provide all the metered pricing guarantees, including guaranteeing backups don't get deleted after a 7 day shot clock. And they may be able to make other guarantees (e.g. priority support) depending on what they can convince Colin to deal with; I'm sure that it's easier to have a single special customer that gets special treatment than it is to offer priority support (and other guarantees) to arbitrary customers.

---

On another note, I disagree with the author that metered pricing should be removed entirely (although he seems to have already conceded that he will never win that argument). Tiered pricing for businesses is great, but there's still a need for cheaper pricing for various reasons.

Perhaps I'm an employee at a company that's testing Tarsnap to ensure it meets the company's needs before making the case for the company to buy a service plan. I may not have the authority to purchase a $50/mo plan, but I certainly would have no qualms about paying $1 out of my own pocket to test the service for two weeks.

Or maybe I'm an independent developer who's bootstrapping a business. I'm still in development mode so I have no customers and therefore no income. Being able to pay $2/mo for the backups I need is a lot more appealing than paying $50/mo. When I finally launch my product and gain users, then I can consider switching to the $50/mo plan for the guarantees it offers (e.g. predictable pricing).

What I love about the concept of metered pricing is how direct it is. You pay exactly for the resources in use. No strange statistics, heuristics and assumptions have been applied to your data, whatever it is, that's what you're using and that's that. You can then look at your month-to-month usage and do the statistics yourself.

I do like the proposals in the article, but I wouldn't make the metered pricing that invisible. I'd call it Elastic Tarsnap, put it in the same table before Tarsnap Professional, give it a faded background and put the language with the picodollar pricing on it. This will communicate that that's the intended product for the truly paranoid power user who needs their personal data backed up. Maybe even put that exact sentence under it "Backups for truly paranoid power users (or unix geeks)". Because it's not just corporations that have specific needs, we unix geeks have specific needs too.