Hacker News new | ask | show | jobs
by pc 5245 days ago
Thanks for the honest feedback.

Also, what is up with all of that downtime? I think I'd expect my payment provider to have better uptime. Makes me wonder how quality the backend really is.

While all downtime is bad, we think 99.99% is decent.

The best service providers -- including the services that might power your site -- tend to have guarantees in this range. Amazon EC2 and Google App Engine both have SLAs for 99.95% uptime.

More generally, this is part of the point: other payments companies don't tend to talk about their availability at all. We think that this is harmful. What should count as acceptable uptime performance is a fair question. As a first step, we think we should all make our uptime public.

Is it just me, or is this really confusing? Which services are responsible for taking payments? If the website goes down, does that affect taking payments? Isn't stripe.js part of the API? If the API goes down, doesn't that affect stripe.js?

No, the Stripe site going down won't impact your ability to accept payments at all. stripe.js is optional, and so we've exposed its uptime separately too. Again, other services like Amazon and Google break out their availability on a per-service basis, and I think this makes a lot of sense.

2 comments

1. The graphic exaggerates your downtime (1 minute/day looks like a day - x1440 worse than it is).

2. (On the site) please explain what each status is and how they relate (or link to rel docs). I understand the website uses the API; yet sometimes the API is down but not the website...

BTW: I humoured some severe criticism of my business once; it was only years later that I realized how accurate, helpful and crucial it was.

Yeah, these are both good points. Thanks.
I don't see a SLA for Stripe on the ToS page, why not?

GAE has actually had 100% uptime for the last year. http://googleappengine.blogspot.com/2012/01/happy-birthday-h...

I don't see a SLA for Stripe on the ToS page, why not?

Blanket SLAs are not normally worth the paper they're written on. Look up the breach-compensation for your favorite TOS-SLA sometime.

Publishing the uptime history is a much more honest approach. Kudos to stripe for doing just that.

I'm not knocking them for publishing their failures. That is all great.

I'm questioning the comparison with other unrelated systems that are much bigger than them, especially when they don't publish a SLA themselves.

Why should they publish an arbitrarily chosen figure in addition to telling you the actual figure?
That is a really good way to put it, but I can't seem to ignore the abscence of an important business document. It feels like the arbitrarily chosen figure is step 1, the actual figure is step 2, and they just skipped the first half.
That blog post refers to their high-replication datastore, not GAE as a whole.