Re: [hybi] name symtax, was: WebSocket sub protocol name.

[Jamie Lokier <[e-mail]>]

Pieter Hintjens wrote:
> First, this breaks in any language where a null is significant in a
> string.  What rules are there for making this work in C?  Or is that
> not considered a valid language for implementors?

Additional language notes:

Java encodes ASCII NUL in UTF-8 with a different encoding to standard
UTF-8.  Therefore, Java and C-like environments do not interoperate
perfectly in the presence of ASCII NUL, unless extra steps are taken -
which are, of course, often ignored.

Unix shells also fail when asked to handle NUL for the most part, due
to their C heritage.

C++ can handle NUL, up until the point you use the string with c_str().
Then it will truncate at the NUL.

Many system protocols, directory services and database queries will
not like being asked to look up something starting with a name
containing NULs and, often because they use C style NUL-terminated
strings in their APIs.

The POSIX regular expression matching functions do not operate
correctly on a string containing embedded NULs.  Unless they are
encoded with Java-UTF-8, that is.

These concerns may apply to the data framing later in the protocol.
When NUL is transmitted in a data frame, either as the byte 0x00, or
the Java-UTF-8 encoding of ASCII NUL, you may expect some
implementations to get confused about the content of a frame.
However, as long as they operate on bytes at the frame boundary
finding stage, boundaries are likely to be found correctly.

-- Jamie