Your phone is the terminal. Stream a Linux/Mac desktop to an iPhone over WebRTC through NAT, with audio and input. Open source, single binaries.
- ..
- people
- README.md 3.6 KB
- apps.sh 6.9 KB
- create-aws.sh 2.3 KB
- create-droplet.sh 3.2 KB
- create-linode.sh 860 bytes
- create-vultr.sh 1.3 KB
- desk.env.example 665 bytes
- dns-park.sh 1.5 KB
- dns-point.sh 1.4 KB
- droplet.sh 6.7 KB
- people.sh 6.2 KB
- provision-host.sh 1.7 KB
- setup-linux.sh 292 bytes
infra
Booting and provisioning hosts. No containers: a plain Ubuntu box, the binaries built from source, and systemd units. The config itself lives in ../provision/ as versioned files; these scripts just create a machine and run the provisioner on it.
How it fits together
create-<provider>.sh boots an Ubuntu 24.04 instance in a region
│
└── provision-host.sh <ip> copies ../provision/ and runs apply.sh
│
└── ../provision/apply.sh installs, configures, starts services
provision/apply.sh is provider-agnostic. Any Ubuntu 24.04 box works, including bare metal or a machine you already have:
infra/provision-host.sh <ip> [host-name] [ssh-user]
Recommended: a domain on DigitalOcean DNS
NC_DOMAIN=nc.example.com infra/create-droplet.sh # droplet + DNS + Let's Encrypt, one command
infra/droplet.sh dns nc.example.com # re-point the name at the current droplet
NC_DOMAIN=nc.example.com infra/droplet.sh provision # turn HTTPS on for an existing box
A rebuilt droplet gets a new IP; with the domain's DNS on DigitalOcean the scripts move the A record (ttl 60) themselves, so the address you hand out never changes and the certificate is always real. dns-point.sh is the building block. Full reasoning and the one-time setup: ../docs/DOMAIN.md.
Providers
Only DigitalOcean has been tested end to end. The others are written the same way and should work, but treat them as first runs.
| Script | Provider | Default region | Notes |
|---|---|---|---|
create-droplet.sh | DigitalOcean | nyc3 | Tested. doctl auth. No central-US region; ~42 ms from Houston. |
create-vultr.sh | Vultr | dfw (Dallas) | vultr-cli + VULTR_API_KEY. Dallas is ~10-20 ms from Houston. |
create-linode.sh | Linode / Akamai | us-central (Dallas) | linode-cli configure. Dallas metro. |
create-aws.sh | AWS EC2 | us-east-2 (Ohio) | aws configure. ~25-30 ms. Opens a security group with the ports. |
Latency is set by geography, not machine type. From Houston, a Dallas box (Vultr/Linode) or an AWS Houston Local Zone wins; DigitalOcean's closest US region is New York.
Getting a key
- DigitalOcean:
doctl auth init(token at cloud.digitalocean.com/account/api/tokens) - Vultr:
export VULTR_API_KEY=...(my.vultr.com/settings/#settingsapi) - Linode:
linode-cli configure(cloud.linode.com/profile/tokens) - AWS:
aws configure(an IAM access key)
Each provider needs your SSH public key registered; the scripts use ~/.ssh/id_ed25519.pub and tell you the one-liner to upload it if it is missing.
AWS Houston Local Zone (lowest latency from Houston)
Local Zones put compute in-metro (single-digit to low-teens ms) but need a bit more plumbing than a standard region. Once aws configure works:
# 1. opt the zone in (Houston is a child of us-east-1)
aws ec2 modify-availability-zone-group --region us-east-1 \
--group-name us-east-1-iah-1 --opt-in-status opted-in
# 2. you then need a subnet in us-east-1-iah-1a with a route to an
# internet gateway, and to launch there. Ask and this can be scripted
# into a create-aws-localzone.sh once you have an account to test on.
Dallas Local Zone is us-east-1-dfw-1 if Houston is unavailable.
DigitalOcean day-to-day
infra/droplet.sh status | ssh | creds | logs | update | provision | dns <name> | off | on | destroy
update pulls and rebuilds the binaries; provision re-copies ../provision/ and re-applies it. Powered-off droplets are still billed; destroy when idle and recreate in about five minutes.