How we'll talk to you when we go down

We will have outages. Anyone who tells you otherwise is either new or lying, and the useful question is not whether a host goes down but what happens in the twenty minutes afterwards.

So here is what we are committing to, in public, before we have anything to be embarrassed about.

What exists today

The status page is public and needs no account. What it does right now is narrow and worth describing precisely rather than dressing up: it probes the platform’s components live, in your browser, and reports each one as operational or degraded. No history, no incident feed, no commentary.

That is a health check, not incident communication. Its one real virtue is that it does not depend on us being awake — if a component stops answering, the page says so whether or not anyone has noticed yet.

If your server is unreachable and the status page says everything is fine, that is itself a bug and we want to hear about it.

What we are committing to add

During an incident, a written update on the status page saying what is affected and what we currently believe is happening. If we do not know yet, the update will say we do not know yet. A holding message with no information still beats silence, because silence is indistinguishable from nobody having noticed.

After an incident, a write-up for anything that took servers offline for a meaningful stretch: what broke, why, and what we changed so it does not happen the same way again. Those will land in the Company category on this blog.

Neither of those has been tested against a real outage yet, because there has not been one to write about. Judge us on the first one.

What we will not do

• Quietly resolve things. If a status entry goes up, a closing note goes with it.
• Call an outage “degraded performance”. If your server was unreachable, the word is down.
• Blame an upstream provider and stop there. Whose hardware it was is context, not an explanation. We chose the provider.
• Make you open a ticket to find out whether it is you or us. That is what the status page is for.

The part that is genuinely hard

We are a small team, not a follow-the-sun operation, and we would rather tell you that than imply a response time we cannot hold at 4am on a Tuesday. The automated check is what covers the gap: it reports a component as degraded on its own, hours before anyone is awake to write a sentence about why.

If you are evaluating hosts, this is worth asking all of them: not “what is your uptime percentage”, which everyone rounds favourably, but “show me your last three incident write-ups”. The absence of any is the answer.