App delivery
This page follows a built frontend as it becomes a live app on the Polkadot
Products Devnet. The path is concrete: publish a static bundle with the pad
CLI, store the content on the Bulletin chain, bind a .dot name to it, and load
it through the dev-dot.li gateway or the Polkadot app.
App delivery is deliberately layered so that no single server sits between a user
and an app. Content lives on-chain (Bulletin), the pointer to it lives on-chain
(a .dot name resolver on Asset Hub), and the loader is a client-side program
that reads both directly.
The two pipelines at a glance
Delivery has a write side (publishing) and a read side (opening). The publisher uploads a bundle and binds a name to it once; every visitor then resolves and fetches that bundle independently.
flowchart TD
A[Built static app] --> B[pad CLI prepares bundle]
B --> C[Upload content to Bulletin]
C --> D[Receive content CID]
D --> E{DotNS on Asset Hub}
E -->|not owned| G[register name]
E -->|owned| H[use existing name]
G --> I[write contenthash]
H --> I
I --> J[optional Browse listing]
J --> K[Live: name.dot in app + https://name.dev-dot.li]
Publishing with pad
The deploy CLI is @parity/polkadot-app-deploy,
which ships the pad binary (alongside polkadot-app-deploy and
polkadot-app-bootstrap). pad selects a network with --env <network>; the
concrete preset name is provided by the team operating the network. After
building your frontend, a publish is a single invocation over the output
directory:
The CLI prepares the static bundle, uploads the content, and binds the resulting
content identifier to the .dot name. Re-publishing the same app updates that
name to point at the new content.
Bulletin storage and upload authorization
Content is stored on the Bulletin chain. Uploads are authorization-based rather than fee-based: the deploy account needs upload quota, but does not pay devnet tokens for each bundle.
Note
Authorizations are finite and expire. If a previously working deploy account starts failing at the upload step, its authorization likely lapsed and must be refreshed before uploads resume.
Binding the .dot name
Once the content CID exists, the CLI checks whether the signer owns the .dot
name. If the name is available, it can register it; if the signer already owns
it, the CLI updates the name's content hash. That single on-chain record is what
turns a name into an app. See Naming (DotNS) for how ownership and
resolution fit together.
Optionally, --publish calls Publisher.publish(label) so directory apps such
as Browse can enumerate your app; it is silently skipped on networks that have
no Publisher contract configured. See App discovery (Browse).
Tip
A deploy config can also publish manifest records. Product apps use those records to describe executables, labels, and icons to the host.
Opening an app through the gateway
The web gateway at https://dev-dot.li is a client-side loader. There is no resolution server: the host shell reads the label from the subdomain, resolves it, fetches the content, and renders it in a sandboxed iframe.
flowchart TD
U[User opens survey.dev-dot.li] --> H[Gateway reads label]
H --> R[Resolve survey.dot]
R --> C[Read contenthash]
C --> D[Decode to CID]
D --> IF[Render app in sandbox]
IF --> FE{Fetch content}
FE -->|Bulletin / light client| V[Verified path]
FE -->|IPFS gateway| T[Gateway path]
V --> RN[App uses Host API bridge]
T --> RN
The rendered app talks to the host through a bridge for accounts, signing, chain
connection, and scoped storage. From the user's perspective, the same name works
in both places: on the web it is https://<label>.dev-dot.li, and in the
Polkadot app it is <label>.dot.
Common blockers
- Upload fails. The deploy account may need Bulletin storage authorization.
- The name cannot be updated. The signer must own the
.dotname. - The app opens but has stale content. Confirm the name's content hash points to the latest CID and that the gateway is resolving the expected network.
- The app is live but hard to find. Use
--publishso Browse can list it.