Getting a remote file's properties with curl, without downloading it
A GET request pulls the whole body. Most of the time I actually want is in the headers — size, type, whether it changed — and there's a way to get each of those without pulling anything.
Headers only: -I
curl -I https://example.com/file.zip
Uses an actual HTTP HEAD request, so the server (in theory) sends the same headers a GET would, minus the body. In practice not every server implements HEAD faithfully — some return a different Content-Length than the real GET would, or skip headers that only get computed while generating the body. If something looks off, cross-check with curl -s -o /dev/null -w (below) against a real GET.
The exact field you want, nothing else: -w/--write-out
Rather than piping -I output through grep, ask curl for one specific value:
curl -s -o /dev/null -w '%{content_type} %{size_download} %{http_code}\n' https://example.com/file.zip
-o /dev/null throws the body away, -s silences the progress meter, and -w prints exactly the fields asked for. Useful ones: content_type, size_download, http_code, time_total, filename_effective (the final URL after redirects, handy when scripting against a link that might bounce somewhere else).
Only fetch it if it changed: -z
curl -z ~/.cache/file.zip -o file.zip https://example.com/file.zip
-z takes either a date string or, as here, an existing local file — curl reads that file's mtime and only downloads if the remote copy is newer (If-Modified-Since under the hood). Prefix the value with - to invert it and match files older than the given time instead of newer.
Keep the remote timestamp after downloading: -R
curl -R -o file.zip https://example.com/file.zip
Sets the downloaded file's local mtime to match the server's Last-Modified header, instead of "just now." Combine it with -z above and repeated runs of the same command become a proper incremental sync: skip if unchanged, and the local timestamp stays accurate when it does download.
Use the server's suggested filename: -J -O
curl -J -O https://example.com/download?id=42
-O alone saves using the URL's own basename — which is useless for a URL like that one. -J tells curl to read the Content-Disposition header instead and use whatever filename the server actually suggests.
Check just a byte range, without the whole file: -r
curl -r 0-99 -o first-100-bytes https://example.com/file.zip
Handy for peeking at a file's format (magic bytes) or confirming a server actually supports partial content (Accept-Ranges: bytes in a -I response) before committing to a resumable download.
Created 2026-09-05T17:19:22+02:00 · Edit