404/403 can be handled with plugins

bin/handlers/caching-proxy.py

@@ -0,0 +1,36 @@
+# assume each requested file exists on another webserver and
+# download + mirror them as they're requested
+# (basically pretend we're warnish)
+
+import os
+import requests
+
+from typing import TYPE_CHECKING
+
+if TYPE_CHECKING:
+    from copyparty.httpcli import HttpCli
+
+
+def main(cli: "HttpCli", vn, rem):
+    url = "https://mirrors.edge.kernel.org/alpine/" + rem
+    abspath = os.path.join(vn.realpath, rem)
+
+    # sneaky trick to preserve a requests-session between downloads
+    # so it doesn't have to spend ages reopening https connections;
+    # luckily we can stash it inside the copyparty client session,
+    # name just has to be definitely unused so "hacapo_req_s" it is
+    req_s = getattr(cli.conn, "hacapo_req_s", None) or requests.Session()
+    setattr(cli.conn, "hacapo_req_s", req_s)
+
+    try:
+        os.makedirs(os.path.dirname(abspath), exist_ok=True)
+        with req_s.get(url, stream=True, timeout=69) as r:
+            r.raise_for_status()
+            with open(abspath, "wb", 64 * 1024) as f:
+                for buf in r.iter_content(chunk_size=64 * 1024):
+                    f.write(buf)
+    except:
+        os.unlink(abspath)
+        return "false"
+
+    return "retry"