24 Dec 1998	Think about canonicalizing values.  This needs a coherent
		think-through.

24 Dec 1998	Aliases for all arrow keys.  Check vi and lynx for some
		good bindings.  Also look at info.

25 Dec 1998	Are there any other hidden environment variables?
		Answer: NO.  Fold this into the "undefined variable" work.

25 Dec 1998	User can delete all the characters in a hex or int and
		the string happily becomes zero length.  Not good.
		Perhaps I need some kind of hook to spring it back
		to "0" when this happens.  This fits in with the whole
		canonicalization issue.  Also ... what happens if someone
		hits left-arrow and right-arrow, left-arrow is going to
		surprise them!  In general, need to be careful about
		navigation commands in a text field.

25 Dec 1998	Choice lists come up without their default.  UPDATE:
		Ok, I have a usable symbol under verb_nchoice.  So now
		this is just part of the general "force defaults"
		problem.

25 Dec 1998	How about a SIGWINCH handler, that would be cool!

25 Dec 1998	Prompt strings need to get truncated at end of line.
		See SCSI low-level driver for example.  Window option or do
		it by hand?

25 Dec 1998	Beef up the symbol table and put in strong (implicit)
		typing!

25 Dec 1998	Get some usability testing on the sixpack (Insert, Delete,
		Home, End, Page Up, Page Down).  I like them, but will anybody
		else?

25 Dec 1998	Parser should validate def_value for hex, int, string.  Update:
		this is hard because sometimes the def_value is a variable.
		Also, I see this idiom: int 'foo' CONFIG_FOO $CONFIG_BAR 10.
		I don't think that idiom is going to work as designed!

25 Dec 1998	Auto indentation (need some if-hooks).

27 Dec 1998	Running on an xterm opens up a can of worms.  My simple
		default configuration has a :li#24: in the termcap entry,
		so I start getting cruft on line 24 of a big screen.
		Then the arrow keys stop working.  Maybe I need to upgrade
		my Red Hat 3.0.4 before getting into this.

27 Dec 1998	More generally, there could be a test on a variable before
		the variable is defined (within the same menu!)  This
		jacks around the aline layout and can leave hline hanging
		out in the breeze.  See x-bug-1.in.

27 Dec 1998	Seed random number generator from /dev/random or /dev/urandom
		(I wonder if I can generate configurations faster than
		/dev/random can build entropy!)

27 Dec 1998	Someone will probably want an option to suppress warnings.
		Sigh.  Fix the warnings, they indicate places where people
		have wrong beliefs in their heads about the semantics of
		Configuration Language.  They can go argue with Configure,
		Menuconfig, and scripts/tkgen.c if they think a boolean
		takes a default value or that any interpreter anywhere
		will handle 'int 'foo' CONFIG_FOO $CONFIG_VALUE 10' in
		the way they want.  (Hey I'm cranky right now!)

27 Dec 1998	Be careful about memory leaks through input_push_string.
		'help' is all right because of the static allocation,
		but 'query' is going to have a problem (unless I am sneaky
		and cache the query menus through the symbol table).

28 Dec 1998	Test new verb_define_int.  Update: also verb_define_hex
		and verb_define_string.

28 Dec 1998	Check out 'make checkhelp', looks like a good thing for
		--mode syntax to do.

29 Dec 1998	The .old file will lose its value on the second 'W' command.

29 Dec 1998	CONFIG_ARCH!  The top-level Makefile does not need to know
		its architecture until after it sources .config.

09 Jan 1999	Careful about statement_bools_allowed on all dep_* statements,
		include hex/int/string.  UPDATE: or maybe just shitcan
		dep_hex, dep_int, dep_string.

09 Jan 1999	Really need input cleaning for 0x on hex.

10 Jan 1999	[Hannu Lyytinen] verb_nchoice, command ' ': if current
		line is not the header line, and not the selected line,
		then select that line rather than cycling.

11 Jan 1999	Add trace facility, specifically for wgetch.

13 Feb 1999	Search for prompt, search for config name.  Implies
		xconfig-like "total display".  E.g. suppose someone
		wants CONFIG_BAR and it depends on CONFIG_FOO and
		CONFIG_EXPERIMENTAL, gotta show that to them.

02 Oct 1999	Generalize test-parser so that language cookie type is
		variable in some way.

11 Oct 1999	New symbol table code looks marginally better.  I think the
		next step is to explicitly walk the tree with scb_something
		in order to set all values when reading a dotconfig.  I like
		the idea of a coherent "value set" -- think about this some
		more.  Also this fits in nicely with Wisdom Of The Penguin.

11 Oct 1999	Have the parser assign symbol types.
