blixtLong fat links
A single TCP stream starves on high-RTT, lossy paths. QUIC with BBR and parallel streams is designed to make better use of long paths. The raw direct-path comparison package is still pending.
Comparison
Blixt vs scp, rsync & SFTP
The infrastructure default, for good reason: free, on every server on earth, scriptable, trusted for decades — and rsync’s delta algorithm is a classic. Blixt is for the transfers where they run out of road: long-RTT links, NAT on both ends, and receivers without shell accounts.
| scp / rsync / SFTP | Blixt | |
|---|---|---|
| Price | Free; ships with the OS | Free plan; flat paid plans |
| Availability | Every server on earth | Windows today; Linux and macOS builds are being finalized — GUI, web UI, and CLI |
| Long-RTT or lossy links | One TCP stream; throughput falls as RTT and loss grow | QUIC + BBR with parallel streams; the raw direct-path comparison package is still pending |
| Short, clean links | Fine — scp saturates a LAN just as well | Same class; nothing to win here |
| NAT traversal | None — needs a reachable host, an open port, keys and accounts | ICE/STUN hole punching plus relays; neither side needs a public IP |
| Sending to a person | They need an account on a machine, or you stage the file on a middle box | They type a code; six digits verify the channel — or they open a share link in a browser |
| Resume after a drop | scp: none. rsync: --partial works but is fragile over flaky links | MAC-protected block journal; only missing chunks are re-sent |
| Delta sync / mirroring | rsync’s home turf — only changed blocks travel | Not what Blixt does — it moves files, not deltas |
| Scriptable | The benchmark for it | Headless CLI — blixt send / blixt recv in cron and CI |
Credit where due
Where scp, rsync & SFTP are the better choice
- Inside infrastructure you already control. Keys deployed, hosts reachable, firewalls yours — nothing is simpler than
scp file host:. - Incremental mirroring and backups.rsync’s delta transfer of changed files is its home turf; Blixt does not compete there.
- Short, clean links. On a LAN or a good same-region path, scp fills the pipe fine. The gap Blixt closes shows up with distance and loss, not across the rack.
- Audits and habits. SSH is universally understood by ops teams and reviewers; that counts for a lot.
And they’re free and open source.OpenSSH and rsync are public, battle-tested code already installed on every box you own; Blixt’s code is not public. For server-to-server work inside your own perimeter, keep using them.
The other direction
Where Blixt wins
No reachable host required
Both ends can sit behind NAT. No port forwarding, no bastion, no creating a shell account for the other side. Hole punching first, relays on :443 when nothing else passes.
Receivers who aren’t ops
A code they type and six digits to compare — in a GUI, a web UI, or a plain browser via a share link. BLAKE3 verifies every chunk end to end.
Scriptable, self-hostable
blixt send dump.sql --code "$CODE" drops into cron and CI like scp does, resumes with the verified chunks still missing after a drop, and blixtd --embed-signal keeps the signaling hub on your own box.