source: trunk/third/vte/README @ 20883

* What is VTE?
- You could say that VTE is something of a research project of mine, based on
  the simple question:  "if programs can use a termcap file (through either
  libtermcap or curses or ncurses) to determine how to drive a terminal, why
  can't a terminal emulator use a termcap file to determine how to behave?"

  Update: the answer is most likely "because applications which use curses
  have more detailed information than that which is found in termcap".

* What does VTE include?
- VTE includes a library (libvte) which implements such a terminal emulator
  widget for GTK+ 2.2/2.4, and a sample application (vte) which wraps that
  widget in a GTK window.  Because I'm more concerned with whether or not it
  works, all settings are hard-coded to whatever I needed to test the last time
  I touched it.  If you actually want to use the widget to get work done, you
  should probably be using gnome-terminal.

* How does it work?
- The VTE library inserts terminal capability strings into a tree of tables,
  and then uses it to determine if data received from a pseudo-terminal is a
  control sequence or just random data.  The sample program "interpret"
  illustrates more or less what the widget sees after it filters incoming data.

* What's missing?
- Accessibility doesn't work quite right yet.
- Mouse hilite tracking isn't implemented yet.  If you'll pardon the pun, for
  certain applications we're terminally unusable without it.
- No support for bidirectional text display.  No support for shaping.  More
  on these topics:
  http://www.opengroup.org/onlinepubs/9638399/overview.htm#tagcjh_03_03_01
  http://www.opengroup.org/onlinepubs/9638399/overview.htm#tagcjh_03_03_02
  http://www.unet.univie.ac.at/aix/aixprggd/genprogc/layout_over.htm
  http://www.ecma-international.org/publications/techreports/E-TR-053.htm
- Most control sequences are recognized, but many aren't implemented.  There
  are enough to run ls, vim, less, emacs and mutt, but more need to be
  implemented (ds, ff, hd, hu, i1, i3, iP, is, LF, LO, MC, ML, mm, mo, pf, pk,
  pl, pn, po, pO, ps, px, r1, r2, r3, RA, RF, rp, rs, RX, SA, SX, wi, several
  more from the XTerm set).
- I'm not sure the widget implementation itself is correct.  There are many
  changes in going from GTK+ 1.2 to 2.x, and examples of the proper way to do
  things is currently scarce, so some of it's guesswork.
- An actual property interface needs to be retrofitted over the various options
  which are only presented as get/set accessors.

* What's weird?
- Relative cursor motion is weird.  When the character to the left of the
  cursor is a 3-byte UTF-8 sequence for a character which occupies two
  columns on the screen, three things may happen when the application sends
  the "cursor left" control sequence:
  * the cursor moves two columns (one character) to the left
    This eliminates the possibility of moving the cursor into the middle of
    a multi-column character.
  * the cursor moves one column (one half-character) to the left
    This makes determining where the cursor is easier, but requires the
    application to emit a control sequence more than once for multi-column
    characters.
  * the cursor moves one "byte" to the left
    This is what most OS kernels do when reading input using "cat".  It
    happens to work for a few locales, and is otherwise just plain broken.
  Currently VTE follows the second convention.  More on this topic:
  http://czyborra.com/unicode/terminals.html
  http://mail.nl.linux.org/linux-utf8/1999-10/msg00014.html
  http://www.debian.org/doc/manuals/intro-i18n/ch-output.en.html
- Depending on your locale, the current encoding, and possibly the strengh
  of cosmic rays, the display width of certain Unicode characters changes.
  For the most part this works correctly now, but if you find that it
  doesn't, please file a bug report including a screenshot, the typescript
  file produced by the incorrectly displayed program run under "script", and
  the contents of your LANG/LC_* environment variables.