chunked multibyte chars problem of mod_websocket

problem detail

This problem occurs in text frame. And a cause of this problem is that mod_websocket has no message format between backend servers.

For example, UTF-8 character encoding for Japanese may be represented by three bytes per character. And mod_websocket uses asynchronous I/O for read/write to backend servers.

In this situation, when mod_websocket receives data from a backend server and mod_websocket reads it devided into 2 bytes and 1 byte, a problem occurs.(Fig.1)

  Fig.1

  +-----------------------+     +-----------------------+
  | browsers              |     | browsers              |
  |                       |     |                       |
  | may show garbled char |     | may show garbled char |
  +-----------------------+     +-----------------------+
              ^                             ^
              |                             |
  +-----------------------+     +-----------------------+
  | Websocket frame       |     | Websocket frame       |
  |                       |     |                       |
  | contains only 2byte   |     | contains only 1byte   |
  +-----------------------+     +-----------------------+
              ^                             ^
              | async write  =>             | async write
  +-----------------------+     +-----------------------+
  | mod_websocket         |     | mod_websocket         |
  |                       |     |                       |
  | construct websocket   |     | construct websocket   |
  | frame w/ only 2bytes  |     | frame w/ only 1byte   |
  +-----------------------+     +-----------------------+
              ^ async read                    async read
              |
  +-----------------------+
  | servers               |
  |                       |
  | send 3bytes           |
  +-----------------------+

way of solving

Sorry, at this time I have no idea to solve this.

It's too difficult to solve this without determining message format between backend server and mod_websocket. (And this means that mod_websocket needs to provide some server-side libraries)

nori0428 / mod_websocket

chunked multibyte chars problem of mod_websocket #11