Design Uploads That Survive a Broken Connection

Forms & Inputs · uploads, loading, errors

Updated 2026-08-04

Send large files in pieces rather than as one transfer. A single stream that fails at ninety percent has to start again, and a chunked one only resends the piece that failed.
Resume rather than restart, and say which one is happening. Resuming from 84% is the difference between a recoverable interruption and an abandoned upload.
Keep the selected file after a failure. Asking someone to find a two gigabyte file again is how an upload becomes a support ticket.
Separate transferring from processing, because they are different waits with different lengths. A video that finishes uploading has not finished being converted, and a progress bar that reaches one hundred percent and then sits there looks stuck.
Design the processing stage on its own terms:

Give every file in a multi-file upload its own progress, status, and retry. One failure must never cancel the other nine.
Validate size and type before the transfer starts. Rejecting a file after four minutes of uploading is the most avoidable failure in this whole flow.
Show a thumbnail or a name and size as confirmation of what was received. Users routinely upload the wrong file and only the interface can catch it.
Handle the duplicate. Uploading the same file twice is common, and detecting it lets you say This file is already here rather than storing it again.

A large upload is a long conversation over an unreliable connection, and the interface earns its keep by never making the user start it over.