* Define a sentinel function, so that the display doesn't get
screwed up the default sentinel when the process finishes.
(And perhaps switch to fundamental or line mode.)

* - term should accept more than one arg, right now it accepts only
    the name of the program to run, this is useful when you want to
    telnet for example and also give the host name to telnet to.hari@sybase.com

- if a command produces otuput, term does not continously scroll,
    instead waits for all output before displaying.

    This is a real bottleneck, sometimes when i want to interrupt a
    command that is producing a lot of output, since eterm buffers the
    output and doesn't spew it out as it gets it(thus not providing
    scrolling), my C-c's don't get trhough fast enough, I have to wait
    till all the output is processed by eterm before being able to
    interrupt the process.

    I believe eterm should *somehow* provide scrolling just like other
    terminal emulators like xterm, that will be really great.hari@sybase.com 

* If cd is called from term-command-hook, and dir is invalid,
just ignore it.  hari@sybase.com

* Performance improvements?  (It is zippy enough on reasonably
fast machines, though.)

* Should more terminal escape sequences be supported?
Yes:  At least the ANSI color escapes (as in colour_xterm).

* The caching variables (e.g. term-current-row) assume that there
is be no random editing of the buffer that we don't know about.  In that
respect, they are somewhat fragile.

* Document the internals better, and make it easier to write
emulators for other kinds of terminals.

* The scrolling behavior is still not quite right.

* Stallman has expressed dislike for the xterm style of switching to the
alternate buffer on the "ti" capability.  An alternative would be to
defer this until the screen is cleared.  That assumes there are programs
that emit ti without following that by a command to erase the window.
I'm not sure there are any such programs (that we care about) ...

*** CHANGES THAT REQUIRE NEW EMACS PRIMITIVES ***

* If ioctl is made accessible to elisp, it may be possible
to support some kind of automatic switching between char mode
and line mode.  It also becomes possible to privide term
primitives to turn off echoing in the inferior.

*** CHANGES TO DO IN OTHER PROGRAMS ***

* In gdb-mode, a keysequence like \C-c\C-s works by sending the command
"step" in gdb-mode.  Emacs goes to a fair bit of trouble to delete the
prompt preceding the command, as well as the echo from the inferior.
In addition to being hairy, it is somewhat fragile (because of possible
type-ahead, and because the inferior might be mixing other output with
the echo).
More robust would be for emacs to send the command "noecho step" (or
"step #noecho") where the "noecho" is an instruction to gdb (readline)
to erase the prompt for this command, and suppress its echo.