WebDAV Module

The WebDAV module registers the webdav remote file system. It implements the unified file-system key schema using only Sync Engine's injected Request abstraction; it has no WebDAV client dependency.

Every request uses HTTP Basic authentication. The module normalizes the endpoint and percent-encodes individual remote path segments before issuing a request.

Supported Backends

The standard mode can theoretically work with any WebDAV service that supports HTTP(S), Basic authentication, ranged GET, PROPFIND, PUT, DELETE, MOVE, and MKCOL. This includes WebDAV installations based on:

• Nextcloud
• ownCloud
• Apache mod_dav
• Microsoft IIS WebDAV
• InfiniCloud
• Koofr WebDAV API
• Synology WebDAV Server
• Seafile's WebDAV extension
• SabreDAV
• Nutstore
• AList, OpenList, and Cloudreve's WebDAV compatibility layer

The optional chunked-upload mode requires Nextcloud's upload protocol and is not standard WebDAV. Enable it only for compatible Nextcloud servers.

Settings

The module registers these settings:

• Server URL: HTTP or HTTPS WebDAV endpoint. Invalid URLs and non-HTTP(S) protocols are rejected; trailing slashes are removed when saved.
• Username: WebDAV account name, trimmed before saving.
• Password: WebDAV account password. Obsidian stores it in its keychain rather than module settings.
• Base directory: Remote directory that becomes the vault root. / uses the endpoint root. Values are trimmed and normalized with a trailing slash. An empty persisted value defaults to <vault name>/ when the module is constructed.
• Use Depth: infinity: Lists a directory tree with one PROPFIND request instead of recursively traversing depth-one responses. It can speed remote discovery, but some servers do not support it. It gives no performance benefit with asymmetric storage enabled.
• Nextcloud-style chunked upload: Uploads large streamed files through Nextcloud's chunked-upload protocol rather than buffering each whole file in memory. Most WebDAV servers do not support this mode.

File Operations

read() sends GET and returns response bytes. write() sends PUT, returning its ETag header when present; otherwise it immediately calls stat() and returns the resulting file UID.

delete() sends DELETE and treats HTTP 404 as already deleted. move() sends MOVE with a Destination header. mkdir() sends MKCOL; recursive creation creates ancestor directories in order and ignores HTTP 405, which indicates an existing directory.

stat() sends a depth-zero PROPFIND for the target, except /, which is returned locally as a folder. The request asks for the resource type, modification time, content length, and ETag. File UIDs use ETag when available, otherwise <mtime>~<size>; folders have no UID. exists() uses the same lookup and returns false for HTTP 404.

PROPFIND response pagination is supported through a Link: <...>; rel="next" response header.

Streaming

readStream() uses ranged GET requests because Request cannot expose a response stream. It requests 2 MiB ranges, keeps at most eight requests in flight, buffers at most eight chunks, and emits chunks in source order. New requests are scheduled only while the stream consumer can accept data. A zero-byte file returns a closed stream.

Without chunked upload, writeStream() collects the complete input stream in memory, then calls write().

With Nextcloud-style chunked upload enabled, it:

1. Creates a UUID-named upload collection under Nextcloud's upload endpoint.
2. Splits the input into 5 MiB numbered chunks and uploads at most three concurrently. Each upload includes Destination and OC-Total-Length headers.
3. Moves the generated .file to the destination.
4. Returns ETag or OC-ETag; when neither is present, it calls stat() for the file UID.

On failure, the module waits for active uploads and deletes the temporary upload collection.

Base Directory Wrapper

WebDAV uses the shared importable prefixWrapper to expose its configured base directory as the file-system root. See file-system wrappers for wrapper-chain details.