3/31/09 KJW -- keithw@mit.edu

This locker includes an internationalized version of libzephyr and the
zephyr clients.

Mainline zephyr doesn't define the character set for displayable
zephyr messages. The Zephyr Programmer's Manual specifies "printable
ASCII form," but this recommendation hasn't been adhered to.

As a result, some clients encode and interpret the message as Latin-1,
and some as UCS in UTF-8. One popular client (BarnOwl) interprets
each field of the zephyr message separately, as UTF-8 if it can, and
Latin-1 otherwise; it sends each field as Latin-1 if it can, and UTF-8
otherwise. The lack of a standardized on-the-wire encoding has caused
a loss of interoperability.

Additionally, no popular zephyr client has been internationalized to
work in different locales and terminal character encodings. zwgc and
zwgc -ttymode directly hand the zephyr message to X11 or to the
terminal without consideration of the terminal's character set.
BarnOwl assumes the user's terminal uses a UTF-8 state machine and
does not work properly outside of a UTF-8 terminal.

The i18n patch in the zdev locker fixed those problems in libzephyr
and the included zephyr clients. It standardizes the on-the-wire encoding
of the zephyr message as UTF-8. It upgrades the minor version number to
ZEPH0.3 to signal this standardized encoding.

Since not every user has a UTF-8 terminal, and Unix programs are
required to function in the portable locale, the zephyr clients now
speak the user's own character encoding (given by the LANG, LC_ALL, or
LC_CTYPE environment variable) and translate to and from the wire
encoding.

At the same time, the i18n patch attempts to gracefully handle zephyrs
in the Latin-1 encoding, both from BarnOwl and from the mainline
zephyr clients in a Latin-1 locale. If an incoming ZEPH0.2 zephyr
can't be parsed as UTF-8, libzephyr silently interprets it as Latin-1
instead. (libzephyr also performs this task for a ZEPH0.3 zephyr, but
since those are supposed to be in UTF-8 always, the library signals
the condition to the user by prepending an "[Invalid incoming
character]" notice to the zephyr message.)

For outgoing zephyrs, libzephyr attempts to interpret the
user-provided text according to the user's own character encoding, and
translates it to UTF-8 for the wire. If the provided text isn't valid
in the user's encoding, libzephyr interprets the user-provided text as
Latin-1, translates it to UTF-8 for the wire, and prepends an "[Invalid
outgoing character]" notice so the recipient will know the message
may have been garbled.