diff --git a/example/README.md b/example/README.md
index 64d3083..6853e51 100644
--- a/example/README.md
+++ b/example/README.md
@@ -72,8 +72,8 @@ if something is missing!
 
 <br>
 
-- [**`w-long-polling`**](w-long-polling#files) &nbsp;&mdash;&nbsp; asynchronous
-  communication without WebSockets.
+- [**`w-long-polling`**](w-long-polling#files) &nbsp;&mdash;&nbsp; old form of
+  asynchronous communication without WebSockets.
 - [**`w-query`**](w-query#files) &nbsp;&mdash;&nbsp; reading URL query
   parameters.
 - [**`w-server-sent-events`**](w-server-sent-events#files) &nbsp;&mdash;&nbsp;
diff --git a/example/w-long-polling/README.md b/example/w-long-polling/README.md
index 62533be..6120adf 100644
--- a/example/w-long-polling/README.md
+++ b/example/w-long-polling/README.md
@@ -2,12 +2,53 @@
 
 <br>
 
-**Next steps:**
+*Long polling* is a technique, largely made obsolete by
+[WebSockets](../k-websocket#files), where a client sends a request, but the
+server does not respond until some data is available. This delayed response
+works as a sort of “push” from the client's point of view, because it comes at
+a time of the server's choosing.
+
+The example implements a long-polling client and server. The core server
+function is `message_loop`:
+
+```ocaml
+let rec message_loop () =
+  let%lwt () = Lwt_unix.sleep (Random.float 2.) in
+  incr last_message;
+
+  let message = string_of_int !last_message in
+  Dream.log "Generated message %s" message;
+
+  begin match !server_state with
+  | Client_waiting f ->
+    server_state := Messages_accumulating [];
+    f message
+  | Messages_accumulating list ->
+    server_state := Messages_accumulating (message::list)
+  end;
+
+  message_loop ()
+```
 
 <br>
 
-[Up to the example index](../#readme)
+This generates a “message” on the server up to every two seconds, to simulate,
+perhaps, other clients sending messages.
 
-<!-- TODO OWASP link; injection general link. -->
-<!-- TODO Link to template syntax reference. -->
-<!-- TODO Link to the right examples section here and from all examples. -->
+If there is already a client waiting for a response, the message is used to
+respond to the client. If not, the message is placed in a list. The next time a
+client sends a request, the messages already in the list are used to respond to
+the client immediately. If a client sends the next request before any more
+messages are generated, the server waits to respond &mdash; hence, “long
+polling.”
+
+<br>
+
+**See also:**
+
+- [**`k-websocket`**](../k-websocket#files) for a more modern way of achieving
+  asynchronous client-server communication.
+
+<br>
+
+[Up to the example index](../#examples)
diff --git a/example/w-long-polling/long_polling.eml.ml b/example/w-long-polling/long_polling.eml.ml
index 4f5aa66..28178ff 100644
--- a/example/w-long-polling/long_polling.eml.ml
+++ b/example/w-long-polling/long_polling.eml.ml
@@ -37,8 +37,10 @@ let last_message =
 let rec message_loop () =
   let%lwt () = Lwt_unix.sleep (Random.float 2.) in
   incr last_message;
+
   let message = string_of_int !last_message in
   Dream.log "Generated message %s" message;
+
   begin match !server_state with
   | Client_waiting f ->
     server_state := Messages_accumulating [];
@@ -46,6 +48,7 @@ let rec message_loop () =
   | Messages_accumulating list ->
     server_state := Messages_accumulating (message::list)
   end;
+
   message_loop ()
 
 let () =