Use attrs internally for the URL preview code & add documentation. (#10753)

docs/development/url_previews.md

@@ -0,0 +1,51 @@
+URL Previews
+============
+
+The `GET /_matrix/media/r0/preview_url` endpoint provides a generic preview API
+for URLs which outputs [Open Graph](https://ogp.me/) responses (with some Matrix
+specific additions).
+
+This does have trade-offs compared to other designs:
+
+* Pros:
+  * Simple and flexible; can be used by any clients at any point
+* Cons:
+  * If each homeserver provides one of these independently, all the HSes in a
+    room may needlessly DoS the target URI
+  * The URL metadata must be stored somewhere, rather than just using Matrix
+    itself to store the media.
+  * Matrix cannot be used to distribute the metadata between homeservers.
+
+When Synapse is asked to preview a URL it does the following:
+
+1. Checks against a URL blacklist (defined as `url_preview_url_blacklist` in the
+   config).
+2. Checks the in-memory cache by URLs and returns the result if it exists. (This
+   is also used to de-duplicate processing of multiple in-flight requests at once.)
+3. Kicks off a background process to generate a preview:
+   1. Checks the database cache by URL and timestamp and returns the result if it
+      has not expired and was successful (a 2xx return code).
+   2. Checks if the URL matches an oEmbed pattern. If it does, fetch the oEmbed
+      response. If this is an image, replace the URL to fetch and continue. If
+      if it is HTML content, use the HTML as the document and continue.
+   3. If it doesn't match an oEmbed pattern, downloads the URL and stores it
+      into a file via the media storage provider and saves the local media
+      metadata.
+   5. If the media is an image:
+      1. Generates thumbnails.
+      2. Generates an Open Graph response based on image properties.
+   6. If the media is HTML:
+      1. Decodes the HTML via the stored file.
+      2. Generates an Open Graph response from the HTML.
+      3. If an image exists in the Open Graph response:
+         1. Downloads the URL and stores it into a file via the media storage
+            provider and saves the local media metadata.
+         2. Generates thumbnails.
+         3. Updates the Open Graph response based on image properties.
+   7. Stores the result in the database cache.
+4. Returns the result.
+
+The in-memory cache expires after 1 hour.
+
+Expired entries in the database cache (and their associated media files) are
+deleted every 10 seconds. The default expiration time is 1 hour from download.