Skip to navigation

Security and data

Where data lives, how long it is kept, how organizations are kept apart and what the status page shows.
View as Markdown
A description, not an audit report

This page describes how the service is built and what it does today. It is not a certification or an audit report. Periods and limits can change, and the changelog lists changes.

Where data lives

The service runs in the AWS region eu-west-1 (Ireland). The status page runs apart, in another EU region, eu-north-1 (Stockholm), so that it stays up when the service is down.

All traffic uses HTTPS. Tillcast has no plain HTTP listener, so a printer set to http:// fails to connect before it sends its secret.

Receipt data

  • Encrypted at rest. The tables of Tillcast are encrypted at rest.
  • Not kept as you sent it. Tillcast renders your document when it accepts the job and keeps only the result, the printer data, for the printer to download. No call returns your document and nothing logs it.
  • Deleted when the job is final. The printer data is deleted when the job is printed, failed, cancelled or expired. A job waits at most 7 days. When a job does not end, for example because the printer has downloaded it and its confirmation never comes, a time limit removes the data from one hour after the job’s expires_at. That clean-up runs in the background and can be late.
  • Not in backups. The printer data lives in a table without point-in-time recovery, so deleted receipts are not kept in backups.
  • Never logged. Receipt documents, rendered output, printer URLs and secrets, API keys and the Authorization header are never written to logs. Neither are request bodies or MAC addresses. Access logs that would record the URL path are off, because the path is the printer’s secret.

Anything free-form that you attach to a job, such as client_reference_id or metadata, is stored with the job for 7 days after it ends, and for a job that never ends until 7 days after its expires_at. Do not put guest data in it.

How long data is kept

Kept today:

DataKept
Rendered receipt dataUntil the job is final. A job that never ends loses it from one hour after its expires_at, which is 7 days and one hour after the job is made at the latest, by a background clean-up that can be late.
Job state, outcome, client_reference_id and metadata7 days after the job is final, and from 7 days after expires_at for a job that never ends, by a background clean-up that can be late
Audit entries of what people do in the dashboard90 days. They are written now, and there is no endpoint or screen to read them yet

Periods that are decided for data that does not exist yet, because the features that make it are planned (events, webhooks, usage statistics and billing):

DataKept
Printer state events: connected, disconnected, condition changes3 months
Lifecycle events: printer created, installed, moved or deleted, and changes to customers and locations1 year
Daily usage statistics90 days
Weekly usage statistics1 year
Monthly statistics, usage records and month closes7 years, counted from the end of the calendar year the month belongs to
Webhook delivery log30 days

When an organization leaves, the rest of its data is to be deleted after 3 months. Closing an organization is done by Tillcast, and its steps are not all built yet. There is no option to keep data longer, not even for a fee.

How organizations are kept apart

  • The organization comes from authentication. For every API call, the organization comes from the API key, or from the verified membership of the signed-in person. It never comes from the request body or URL.
  • Another organization’s objects look missing. The API answers “not found”, never “forbidden”, so nobody can learn what exists in another organization.
  • Keys can be limited to one customer. Such a key sees and changes only that customer.
  • A printer sees only its own jobs. Tillcast finds the printer from its secret, and takes everything else from its own records, whatever job id the printer sends.
  • Customer codes grant nothing. A customer code is public. Without a valid secret a request is refused.
  • Secrets are hashed. Tillcast keeps only hashes of printer secrets and API keys. A full URL or key is shown once, when it is issued.
  • A printer address can name one allowed device. This is an optional safeguard on top of the secret. See Printer URL.
  • Browsers. The API accepts calls from a browser only from the dashboard. Call it from your server, and never put a key in a web page.

Sign-in and roles

People sign in to the dashboard with an e-mail address and password, and can add an authenticator app or a passkey. A person can belong to several organizations and has one role in each: owner, developer or viewer. Sensitive actions, such as creating an API key or changing members, need an extra confirmation by a code that is e-mailed to the person. See Authentication and API keys.

Status page

The status page is live at status.tillcast.com. See Status and incidents. It shows these services, each with its state and its availability over the last 7 days by the hour and over the last 30 and 90 days by the day:

  • Printer connections
  • Printing
  • Public API
  • Dashboard

Webhooks are planned and do not exist yet; a webhooks service joins the page when they do. Availability is measured by a canary: every 5 minutes a test job goes through the whole path and is confirmed like a real printer would. The page also lists current and past incidents with their updates.

Report a problem

To report a security problem, write to contact@tillcast.com.