Symbolic DSN file	             decide syntax
i18n (gettext)                       provide framework
Multiple Password attempts           Can I be arsed?
More detailed help files             Farm this out to users, maybe?
				     Too much work for just me, plus I'm not 
				     necessarily using all the dbs I write 
				     drivers for. Hmm. Just tried to use sf's 
				     task system. I think it'll be easier for 
				     me to write all the docs myself after all.
Hooks for script language engines    design this, already
Gtk GUI                              IP (hammer out basic arch)

############################################################################

Init files [global & per-driver]               Done
Paged output				       Done
Editor functionality			       Done
xterm/rxvt titles			       Done ( bugs in various terminal
                                               emulators: rxvt, gnome-terminal,
					       patches sent to maintainers )
Bash style multi->single line history          Done
Zsh style uniq'ing of history                  Done
Controllable EOL character		       Done, GRL bug notwithstanding
                                               Woohoo.

############################################################################

ii8n: shouldn't be too hard to add the framework for this - I just want a 
tiny bit more flexibilty than raw gettext, a thin DBIShell::i18n wrapper
should suffice.

############################################################################

Implement following EOL rules/behaviour:

1) History file _always_ has default '/' terminator in it.

a) If EOL is unset, do nothing special, should all just work
b) If EOL is set, when asked to rl_<fetch history in some way> [RLFETCH],
   do the following:
   
   1) IF CEOL (strip CEOL, attach DEOL)
   2) RLFETCH
   3) IF DEOL (strip DEOL, attach CEOL)

Ok - this next bit would seem not to be rquired: it's only when we fetch a 
line from the TRL_GNU history and then alter and discard it _without_ using 
it that the recorded line gets altered.

# Additionally, if we are exec'ing a command, then _after_ we determine
# is_terminated, we must (for TRL_GNU at least) use readline methods to
# restore DEOL/remove CEOL from the readline private buffer, or the 
# history will be munged. (TRL_GNU appears to actually edit a history line 
# when you mangle a fetched line, rather than a copy of said line). Am 
# uncertain as to whether this is a bug (docs are vague), but it is 
# annoying.