Wirebook › Bonus lesson

Why can a download pick up where it left off?

Bonus lesson · about 6 minutes · hands-on · nothing you type leaves your browser

You start downloading a big file, the train goes into a tunnel, and the connection drops. When you try again, does the computer start from scratch — or ask for only the missing pieces? Here you can break a pretend download into numbered pieces, interrupt it, and see why “resume” can skip the bits you already have.

Numbered pieces of a pretend file

This download lives only in your browser — nothing is uploaded. Press Start, then Interrupt. Try again with restart from zero versus resume with ranges and watch the piece counters.

Restart from piece 1 every time Resume: ask only for missing pieces

0pieces kept

12still needed

0pieces sent this run

0re-sent (wasted)

What the pieces are really called

A big download is just a long row of bytes. When the connection dies at byte 400,000, your device already has the first 400,000. Asking for the whole file again would waste time and data. Instead it can send a range request: “give me from byte 400,000 to the end.” The server replies with status 206 and only those bytes. Your device stitches the old piece and the new piece into one complete file.

Number the slices. Each piece has a start and end byte, like chapter numbers in a book.
Keep what arrived. Saved pieces stay on disk (or in memory) until the file is whole.
Ask only for gaps. A resume request names the missing ranges; a restart throws the kept pieces away.
Check the glue. Many downloads also check a fingerprint (checksum) so a damaged piece is fetched again, not trusted.

When resume works — and when it doesn’t

Situation	What usually happens
Server supports ranges (Accept-Ranges: bytes)	Resume asks for missing bytes; 206 comes back with just that slice
Server ignores ranges	You get the whole file again (200 OK) — restart behaviour
File changed on the server mid-download	Old pieces may no longer match; a careful client starts over
Tiny file / first visit	Often just fetch the whole thing once — resume is for long transfers

It's a bit like… collecting a sticker album. If the bus jolts and you drop half the pack, you don’t buy every sticker again — you note the numbers you still need and ask only for those. The album page keeps the ones that already stuck.

Related lessons: why a video call goes blurry (squeezing a live stream), how a message crosses the world (packets), how a game remembers progress (keeping work on disk). Grown-up twin: Anvil’s HTTP status helpers, and Tooltern guides on HTTP 206 Partial Content, when a download won’t resume, 416 Range Not Satisfiable, resuming with curl or wget, reading the Content-Range label, plus ETIMEDOUT / ECONNRESET when a transfer dies mid-flight.

Words from this lesson

Earn a bonus badge

Get all 3 right to collect the Interrupted downloads bonus badge. Bonus badges are extra: they don't change your stamp count or rank. Wrong answers are fine: just try again.

1. A big download stops halfway. What does a clever resume do next? Delete everything and pretend it never started Keep the pieces it already has and ask only for the missing numbered ranges Download the file twice as fast by magic

2. What does HTTP status 206 Partial Content mean? The whole website is broken The server is sending only the byte range you asked for, not the entire file The file is encrypted and cannot be read

3. Why is restart-from-zero wasteful after an interruption? Because trains dislike tunnels Because bytes you already saved get fetched again Because files can only download at night

Why can a download pick up where it left off?

Numbered pieces of a pretend file

Ask for a numbered range of bytes

What the pieces are really called

When resume works — and when it doesn’t

Words from this lesson

Earn a bonus badge

Try another lesson