TODO 58.4 KB
Newer Older
1 2
Emacs TODO List                                                   -*-outline-*-

Paul Eggert's avatar
Paul Eggert committed
Copyright (C) 2001-2019 Free Software Foundation, Inc.
4 5
See the end of the file for license conditions.

Gerd Moellmann's avatar
Gerd Moellmann committed

7 8 9
If you are ready to start working on any of these TODO items, we
appreciate your help; please write to so we can be
aware that the problem is being addressed, and talk with you how to do
10 11 12 13 14 15
it best.  Also to check that it hasn't been done already, since we
don't always remember to update this file!  It is best to consult
the latest version of this file in the Emacs source code repository.

Since Emacs is an FSF-copyrighted package, please be prepared to sign
legal papers to transfer the copyright on your work to the FSF.
16 17 18 19 20 21
Copyright assignment is a simple process.  Residents of some countries
can do it entirely electronically.  We can help you get started, and
answer any questions you may have (or point you to the people with the
answers), at the mailing list.

For more information about getting involved, see the CONTRIBUTE file.
22 23

As well as the issues listed here, there are bug reports at
<>.  Bugs tagged "easy" ought to be suitable for
25 26 27 28 29 30
beginners to work on, but unfortunately we are not very good at using
this tag.  Bugs tagged "help" are ones where assistance is required,
but may be difficult to fix.  Bugs with severity "important" or higher
are the ones we consider more important, but these also may be
difficult to fix.  Bugs with severity "minor" may be simpler, but this
is not always true.

32 33 34 35 36 37 38 39 40 41 42 43 44 45
* Speed up Elisp execution
** Speed up function calls
Change src/bytecode.c so that calls from byte-code functions to byte-code
functions don't go through Ffuncall/funcall_lambda/exec_byte_code but instead
stay within exec_byte_code.

** Improve the byte-compiler to recognize immutable (lexical) bindings
and get rid of them if they're used only once and/or they're bound to
a constant expression.

Such things aren't present in hand-written code, but macro expansion and
defsubst can often end up generating things like
(funcall (lambda (arg) (body)) actual) which then get optimized to
(let ((arg actual)) (body)) but should additionally get optimized further
when 'actual' is a constant/copyable expression.
47 48 49 50 51 52 53 54 55

** Add an "indirect goto" byte-code and use it for local lambda expressions.
E.g. when you have code like

   (let ((foo (lambda (x) bar)))
      (funcall foo toto)
      (blabla (funcall foo titi))))

turn those 'funcalls' into jumps and their return into indirect jumps back.
57 58 59 60 61 62 63 64 65 66 67 68 69

** Compile efficiently local recursive functions

Similar to the previous point, we should be able to handle something like

   (letrec ((loop () (blabla) (if (toto) (loop))))

which ideally should generate the same byte-code as

   (while (progn (blabla) (toto)))

* Things that were planned for Emacs-24
70 71

** concurrency: including it as an "experimental" compile-time option
72 73
  sounds good.  Of course there might still be big questions around "which form
  of concurrency" we'll want.
** better support for dynamic embedded graphics: I like this idea (my
Glenn Morris's avatar
Glenn Morris committed
  mpc.el code could use it for the volume widget), though I wonder if the
76 77 78 79 80 81 82 83 84
  resulting efficiency will be sufficient.
** Spread Semantic.
** Improve the "code snippets" support: consolidate skeleton.el, tempo.el,
  and expand.el (any other?) and then advertise/use/improve it.
** Improve VC: yes, there's a lot of work to be done there :-(

** Random things that cross my mind right now that I'd like to see (some of
them from my local hacks), but it's not obvious at all whether they'll
make it.
Glenn Morris's avatar
Glenn Morris committed
85 86
*** prog-mode could/should provide a better fill-paragraph default
  that uses syntax-tables to recognize string/comment boundaries.
87 88 89 90 91 92 93
*** provide more completion-at-point-functions.  Make existing
  in-buffer completion use completion-at-point.
*** "functional" function-key-map that would make it easy to add (and
  remove) mappings like "FOO-mouse-4 -> FOO-scroll-down",
  "FOO-tab -> ?\FOO-\t", "uppercase -> lowercase", "[fringe KEY...] ->
  [KEY]", "H-FOO -> M-FOO", "C-x C-y FOO -> H-FOO", ...

Glenn Morris's avatar
Glenn Morris committed
94 95 96 97
* Things related to

** Move idlwave to
Need to sync up the Emacs and external versions.
See <>

Glenn Morris's avatar
Glenn Morris committed
** Move Org mode to
101 102
See <>
Glenn Morris's avatar
Glenn Morris committed
103 104

** Move verilog-mode to
See <>
Glenn Morris's avatar
Glenn Morris committed
106 107

** Move vhdl-mode to
See <>
Glenn Morris's avatar
Glenn Morris committed

* Simple tasks. These don't require much Emacs knowledge, they are
111 112
suitable for anyone from beginners to experts.

113 114
** Convert modes that use view-mode to be derived from special-mode instead.

Eli Zaretskii's avatar
Eli Zaretskii committed
** Major modes should have a menu entry.

Dan Nicolaescu's avatar
Dan Nicolaescu committed
** Check if all items on the mode-line have a suitable tooltip for all modes.

Dan Nicolaescu's avatar
Dan Nicolaescu committed
119 120 121
** edebug and debugger-mode should have a toolbar.
It can use the same icons as gud.

122 123 124
** Check what minor modes don't use define-minor-mode and convert them
to use it.

** Convert all defvars with leading '*' in the doc-strings into defcustoms
Glenn Morris's avatar
Glenn Morris committed
126 127
of appropriate :type and :group.

** Remove any leading '*'s from defcustom doc-strings.
[done?] [A lot of them are in CC Mode.]
Glenn Morris's avatar
Glenn Morris committed
130 131 132 133 134 135

** Remove unnecessary autoload cookies from defcustoms.
This needs a bit of care, since often people have become used to
expecting such variables to always be defined, eg when they modify
things in their .emacs.

136 137
** See if other files can use generated-autoload-file (see eg ps-print).

138 139 140 141
** Write more tests.  Pick a fixed bug from the database, write a test
case to make sure it stays fixed.  Or pick your favorite programming
major-mode, and write a test for its indentation.  Or a version
control backend, and write a test for its status parser.  Etc.
See the 'test' directory for examples.

* Small but important fixes needed in existing features:

146 147 148 149 150 151
** Flymake's customization mechanism needs to be both simpler (fewer
levels of indirection) and better documented, so it is easier to
understand.  I find it quite hard to figure out what compilation
command it will use.

I suggest totally rewriting that part of Flymake, using the simplest
Paul Eggert's avatar
Paul Eggert committed
mechanism that suffices for the specific needs.  That will be easy
153 154
for users to customize.

Kim F. Storm's avatar
Kim F. Storm committed
155 156 157
** Distribute a bar cursor of width > 1 evenly between the two glyphs
   on each side of the bar (what to do at the edges?).

** revert-buffer should eliminate overlays and the mark.
Martin Rudalics's avatar
Martin Rudalics committed
   For related problems consult the thread starting with

162 163
** erase-buffer should perhaps disregard read-only properties of text.

Stefan Monnier's avatar
Stefan Monnier committed
** Fix the kill/yank treatment of invisible text.  At the moment,
Glenn Morris's avatar
Glenn Morris committed
  invisible text is placed in the kill-ring, so that the contents of
Stefan Monnier's avatar
Stefan Monnier committed
166 167
  the ring may not correspond to the text as displayed to the user.
  It ought to be possible to omit text which is invisible (due to a
Glenn Morris's avatar
Glenn Morris committed
168 169
  text-property, overlay, or selective display) from the kill-ring.

170 171 172
** Feature to change cursor shape when Emacs is idle (for more than
  a specified time).

173 174 175
** The buttons at the top of a custom buffer should not omit
  variables whose values are currently hidden.

176 177 178 179 180
** Clean up the variables in browse-url.  Perhaps use a shell command string to
  specify the browser instead of the mushrooming set of functions.
  See also ESR's proposal for a BROWSER environment variable

181 182
** Enhance scroll-bar to handle tall line (similar to line-move).

183 184 185 186
** In Custom buffers, put the option that turns a mode on or off first,
  using a heuristic of some kind?

** Define recompute-arg and recompute-arg-if for fix_command to use.
Eli Zaretskii's avatar
Eli Zaretskii committed
  See rms message of 11 Dec 05 in
Eli Zaretskii's avatar
Eli Zaretskii committed
  and the rest of that discussion.
190 191 192 193

** In Emacs Info, examples of using Customize should be clickable
   and they should create Custom buffers.

194 195
** The toolbar should show keyboard equivalents in its tooltips.

196 197
** Add function to redraw the tool bar.

198 199 200
** Redesign the load-history data structure so it can cope better
  with evaluating definitions of the same function from different files,
  recording which file the latest definition came from.
Kim F. Storm's avatar
Kim F. Storm committed

202 203
** make back_comment use syntax-ppss or equivalent.

Glenn Morris's avatar
Glenn Morris committed
** Consider improving src/sysdep.c's search for a fqdn.
Glenn Morris's avatar
Glenn Morris committed
206 207

** Find a proper fix for rcirc multiline nick adding.
Glenn Morris's avatar
Glenn Morris committed

210 211
** Check for any included packages that define obsolete bug-reporting commands.
Change them to use report-emacs-bug.
212 213 214 215 216 217 218 219
*** Related functions:
**** gnus-bug
**** report-calc-bug
**** org-submit-bug-report
**** lm-report-bug
**** tramp-bug
**** c-submit-bug-report
[Do all of them need changing?]

Nick Roberts's avatar
Nick Roberts committed
221 222
** Allow fringe indicators to display a tooltip (provide a help-echo property?)

223 224 225
** Add a defcustom that supplies a function to name numeric backup files,
like make-backup-file-name-function for non-numeric backup files.

226 227
** 'dired-mode' should specify the semantics of 'buffer-modified-p' for
dired buffers and DTRT WRT 'auto-revert-mode'.
Martin Rudalics's avatar
Martin Rudalics committed

Glenn Morris's avatar
Glenn Morris committed
** Check uses of prin1 for error-handling.
Glenn Morris's avatar
Glenn Morris committed

* Important features:
Glenn Morris's avatar
Glenn Morris committed
233 234

** "Emacs as word processor"
Glenn Morris's avatar
Glenn Morris committed
    rms writes:
Paul Eggert's avatar
Paul Eggert committed
    25 years ago I hoped we would extend Emacs to do WYSIWYG word
Glenn Morris's avatar
Glenn Morris committed
238 239 240
    processing.  That is why we added text properties and variable
    width fonts.  However, more features are still needed to achieve this.

** Extend text-properties and overlays
242 243 244
*** Several text-property planes
This would get us rid of font-lock-face property (and I'd be happy to
get rid of char-property-alias-alist as well) since font-lock would
simply use the 'face' property in the 'font-lock' plane.

Basically 'put-text-property' and friends would take an extra argument PLANE
248 249 250 251 252
(maybe the best backward-compatible way to do that is to make it so that
PROPERTY can be a cons cell (PLANE . PROP)).  So font-lock would
do (put-text-property start end '(font-lock . face) value).

All the properties coming from the various planes would get merged via an Elisp
function (so it can merge 'face' differently than 'keymap' or it could give
254 255
different priorities to different planes (we could imagine enabling/disabling
planes)).  The merging would not happen lazily while looking up properties but
instead it would take place eagerly in 'add-text-properties'.  This is based on
257 258 259 260 261 262 263 264 265 266
the idea that it's much more frequent to lookup properties than to
modify them.  Also, when properties are looked up during redisplay, we
generally can't run Elisp code, whereas we generally can do that when
properties are added.

*** Move overlays to intervals.c

Currently overlays are implemented as (two) sorted singly linked lists (one
for overlays_before some position and one for overlay_after that
position, for some quirky definition of "before" and "after").
The function 'overlay-recenter' changes the position used for the split
268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297
(and is called internally in various situations).

Each overlay is itself implemented with two markers (which keep track of
the overlay-start and overlay-end).  Markers are implemented as
a non-sorted singly linked list of markers.  So every text
insertion/deletion requires O(N) time, where N is the number of markers
since we have to go down that list to update those markers that are
affected by the modification.

You can start in src/buffer.[ch], maybe grepping for overlays_before for
a starting point.

Text-properties, OTOH, are implemented with a (mostly) balanced binary
tree.  This is implemented in src/intervals.[ch].

So we'd like to change overlays so that they don't use markers (and we
don't keep them in two sorted singly-linked lists) any more.  Instead,
we'll store them inside the balanced binary tree used for
text-properties.  I think we can use the "augmented tree" approach
described in

To ease up debugging during development, I'd guess the implementation
would first add the new stuff, keeping the old stuff (i.e. add to
Lisp_Overlay whichever fields are needed for the new code, while keeping
the old ones, add needed overlay fields to the intervals tree, but keep
the old fields, the overlays_before etc...).  This way, you can add
consistency checks that make sure the new code computes the same results
as the old code.  And once that works well, we can remove the old code
and old fields.

298 299 300 301 302 303 304 305 306 307 308 309 310 311
** Having tabs above a window to switch buffers in it.

** "Perspectives" are named persistent window configurations.  We have
had the window configuration mechanism in GNU Emacs since the
beginning but we have never developed a good user interface to take
advantage of them.  Eclipse's user interface seems to be good.

Perspectives work well even if you do the equivalent of C-x 4 C-f
because of the distinction between view windows vs file windows.  In
Emacs this is more or less the "dedicated window" feature, but we have
never really made it work for this.

Perspectives also need to interact with the tabs.

Glenn Morris's avatar
Glenn Morris committed
** FFI (foreign function interface)
See eg
Glenn Morris's avatar
Glenn Morris committed

315 316 317 318
One way of doing this is to start with fx's dynamic loading, and use it
to implement things like auto-loaded buffer parsers and database
access in cases which need more than Lisp.

Glenn Morris's avatar
Glenn Morris committed
** Replace unexec with a more portable form of dumping
320 321
See eg
322 323

One way is to provide portable undumping using mmap (per gerd design).
Glenn Morris's avatar
Glenn Morris committed

325 326 327
** Imenu could be extended into a file-structure browsing mechanism
using code like that of customize-groups.

** Display something in the margin on lines that have compilation errors.
329 330 331 332 333 334 335

** Compilation error navigation bar, parallel to the scroll bar,
indicating where in the buffer there are compilation errors.
Perhaps we could arrange to display these error indications on top
of the scroll bar itself.  That depends on to what extent toolkit
scroll bars are extensible.

** Provide user-friendly ways to list all available font families,
Juanma Barranquero's avatar
Juanma Barranquero committed
  list fonts, display a font as a sample, etc.  [fx is looking at
  multilingual font selection for the Unicode branch of Emacs.]

340 341
** Provide a convenient way to select a color with the mouse.

** Rewrite the face code to be simpler, clearer and faster.

** Program Enriched mode to read and save in RTF.  [Is there actually a
Dave Love's avatar
Dave Love committed
  decent single definition of RTF?  Maybe see info at
Eli Zaretskii's avatar
Eli Zaretskii committed
346]  This task seems to be addressed
  by, which is still in
Eli Zaretskii's avatar
Eli Zaretskii committed
  very early stages.

350 351 352 353 354 355
  Another place to look is the Wikipedia article at

  It currently points to the latest spec of RTF v1.9.1 at

** Implement primitive and higher-level functions to allow filling
357 358
  properly with variable-pitch faces.

** Implement intelligent search/replace, going beyond query-replace
Chong Yidong's avatar
Chong Yidong committed

362 363 364
** Implement other text formatting properties.
*** Footnotes that can appear either in place or at the end of the page.
*** text property that says "don't break line in middle of this".
365 366
   Don't break the line between two characters that have the
   same value of this property.
*** Discretionary hyphens that are not visible when they are at end of line.

369 370
** Internationalize Emacs's messages.

371 372 373
** Set up a facility to save backtraces when errors happen during
specified filters, specified timers, and specified hooks.

374 375
** Install's no-flicker change.

376 377 378 379
** Add a "current vertical pixel level" value that goes with point,
  so that motion commands can also move through tall images.
  This value would be to point as window-vscroll is to window-start.

380 381
** Address internationalization of symbols names essentially
  as documentation, e.g. in command names and Custom.

** Make the Lucid menu widget display multilingual text.  [This
Dave Love's avatar
Dave Love committed
384 385
  probably needs to be done from actual Emacs buffers, either directly
  in the menu or by rendering in an unmapped window and copying the
Stefan Monnier's avatar
Stefan Monnier committed
386 387
  pixels.  The current code assumes a specific locale; that isn't good
  enough even if X can render the arbitrary text] [The gtk
Dave Love's avatar
Dave Love committed
388 389
  port now displays multilingual text in menus, but only insofar as
  Emacs can encode it as utf-8 and gtk can display the result.]
Stefan Monnier's avatar
Stefan Monnier committed
390 391
  Maybe making Lucid menus work like Gtk's (i.e. just force utf-8) is good
  enough now that Emacs can encode most chars into utf-8.

393 394 395
** The GNUstep port needs some serious attention, ideally from someone
familiar with GNUstep and Objective C.

396 397
* Other features we would like:

398 399 400
** A more modern printing interface.  One that pops up a dialog that lets
you choose printer, page style, etc.
Integration with the Gtk print dialog is apparently difficult.  See eg:
401 402

Stefan Monnier's avatar
Stefan Monnier committed
404 405
** Allow frames(terminals) created by emacsclient to inherit their environment
from the emacsclient process.

407 408
** Give Tar mode all the features of Archive mode.

** Create a category of errors called 'process-error'
410 411
  for some or all errors associated with using subprocesses.

** Maybe reinterpret 'parse-error' as a category of errors
413 414
  and put some other errors under it.

415 416
** Make byte-compile warn when a doc string is too wide.

417 418
** Make byte-optimization warnings issue accurate line numbers.

419 420 421 422 423 424 425
** Record the sxhash of the default value for customized variables
  and notify the user (maybe by adding a menu item or toolbar button,
  as the detection can occur during autoload time) when the default
  changes (meaning that new versions of the Lisp source with a changed
  default value got installed) and offer ediff on the respective
  customization buffers.

** Emacs Lisp mode could put an overlay on the defun for every
  function that has advice.  The overlay could have 'after-text' like
428 429 430
  " [Function has advice]".  It might look like (defun foo [Function
  has advice] (x y) The overlay could also be a button that you could
  use to view the advice.

432 433
** Add a function to get the insertion-type of the markers in an overlay.

Stefan Monnier's avatar
Stefan Monnier committed
434 435
** ange-ftp
*** understand sftp
   This is hard to make work because sftp doesn't print status messages.
Eli Zaretskii's avatar
Eli Zaretskii committed

Stefan Monnier's avatar
Stefan Monnier committed
438 439
*** Use MLS for ange-ftp-insert-directory if a list of files is specified.

440 441 442 443
** Ability to map a key, including all modified-combinations.
   E.g map mouse-4 to wheel-up as well as M-mouse-4 -> M-wheel-up
   M-C-mouse-4 -> M-C-wheel-up, H-S-C-M-s-double-mouse-4 ->
   H-S-C-M-s-double-wheel-up, ...

** Beefed-up syntax-tables.
*** recognize multi-character syntactic entities like 'begin' and 'end'.
Juanma Barranquero's avatar
Juanma Barranquero committed
*** nested string-delimiters (for PostScript's (foo(bar)baz) strings).
448 449 450 451 452 453 454
*** support for infix operators (with precedence).
*** support for the $ (paired delimiter) in parse-partial-sexp.
*** support for hook-chars whose effect on the parsing-state is specified
	by elisp code.  Thus a char could both close a string and open a comment
	at the same time and do it in a context-sensitive way.
*** ability to add mode-specific data to the partial-parse-state.

455 456
** Add a way to convert a keyboard macro to equivalent Lisp code.

** Have a command suggestion help system that recognizes patterns
458 459
  of commands which could be replaced with a simpler common command.
  It should not make more than one suggestion per 10 minutes.

** Add a way to define input methods by computing them (when first used)
462 463 464 465
  from other input methods.  Then redefine C-x 8 to use a
  user-selected input method, with the default being the union of
  latin-1-prefix and latin-1-postfix.

** Implement a clean way to use different major modes for
467 468
  different parts of a buffer.  This could be useful in editing
  Bison input files, for instance, or other kinds of text
Eli Zaretskii's avatar
Eli Zaretskii committed
469 470
  where one language is embedded in another language.  See and also
Glenn Morris's avatar
Glenn Morris committed
  mmm-mode, as reference for approaches taken by others.
Dave Love's avatar
Dave Love committed

473 474 475
** Arrange a way for an input method to return the first character
  immediately, then replace it later.  So that C-s a with
  input method latin-1-postfix would immediately search for an a.
Dave Love's avatar
Dave Love committed

** Give start-process the ability to direct standard-error
Dave Love's avatar
Dave Love committed
478 479
  output to a different filter.

** Give desktop.el a feature to switch between different named desktops.

482 483
** Add a cpio mode, more or less like tar mode.

484 485
** Save undo information in special temporary files, and reload it
  when needed for undoing.  This could extend undo capacity.
486 487
  undo-tree, in ELPA, already does this; its saving code could be
  integrated without requiring the use of undo-tree.
Dave Love's avatar
Dave Love committed

** Change the Windows NT menu code
Dave Love's avatar
Dave Love committed
490 491 492 493 494 495
  so that it handles the deep_p argument and avoids
  regenerating the whole menu bar menu tree except
  when the user tries to use the menubar.

  This requires the RIT to forward the WM_INITMENU message to
  the main thread, and not return from that message until the main
  thread has processed the MENU_BAR_ACTIVATE_EVENT and regenerated
Dave Love's avatar
Dave Love committed
  the whole menu bar.  In the mean time, it should process other messages.
Gerd Moellmann's avatar
Gerd Moellmann committed

** Get some major packages installed: W3 (development version needs
Stefan Monnier's avatar
Stefan Monnier committed
  significant work), PSGML, _possibly_ ECB.
Glenn Morris's avatar
Glenn Morris committed
502 503 504
  Check the assignments file for other packages which might go in and
  have been missed.

505 506 507 508
** Make compiler warnings about functions that might be undefined at run time
 smarter, so that they know which files are required by the file being
 compiled and don't warn about functions defined in them.

Glenn Morris's avatar
Glenn Morris committed
** Split out parts of lisp.h.
Dave Love's avatar
Dave Love committed

** Update the FAQ.
Dave Love's avatar
Dave Love committed

** Allow auto-compression-mode to use zlib calls if zlib is available.
Dave Love's avatar
Dave Love committed
  [It's required for PNG, so may be linked anyhow.]
Dave Love's avatar
Dave Love committed

** Improve the GC (generational, incremental).  (We may be able to use
  the Boehm collector.)  [See the Boehm-GC branch in CVS for work on this.]
Gerd Moellmann's avatar
Gerd Moellmann committed

** Check what hooks would help Emacspeak -- see the defadvising in W3.
Dave Love's avatar
Dave Love committed

521 522
** Add definitions for symbol properties, for documentation purposes.

Martin Rudalics's avatar
Martin Rudalics committed
523 524
** Temporarily remove scroll bars when they are not needed, typically
  when a buffer can be fully displayed in its window.
Dave Love's avatar
Dave Love committed

** Provide an optional feature which computes a scroll bar slider's
  size and its position from lines instead of characters.
Dave Love's avatar
Dave Love committed

** Allow unknown image types to be rendered via an external program
Glenn Morris's avatar
Glenn Morris committed
  converting them to, say, PBM (in the same way as PostScript?). [does
Glenn Morris's avatar
Glenn Morris committed
531 532
  doc-view.el do this, or could it be extended to do this?
  Does ImageMagick obsolete this idea?]

** Allow displaying an X window from an external program in a buffer,
535 536 537
  e.g. to render graphics from Java applets.  [gerd and/or wmperry
  thought this was feasible.]

** Allow images (not just text) in the margin to be mouse-sensitive.
539 540 541
  (Requires recursing through display properties).  Provide some way
  to simulate mouse-clicks on marginal text without a mouse.

** Extend ps-print to deal with multiple font sizes, images, and extra
543 544

** Use the XIE X extension, if available, for image display.
Gerd Moellmann's avatar
Gerd Moellmann committed

** Make monochrome images display using the foreground and background
548 549
  colors of the applicable faces.

** Make 'format-time-string' preserve text properties like 'format'.

552 553 554
** Optionally make the cursor a little thinner at the end of a line
  or the end of the buffer.

** Port the conservative stack marking code of Emacs's garbage collector
Eli Zaretskii's avatar
Eli Zaretskii committed
556 557
  to more systems, so that we can completely get rid of GCPROs.  Note
  that Boehm garbage collector provides this.

** Reorder defcustom's in each package so that the more important
560 561
  options come first in the Customize buffers.  This could be done by
  either rearranging the file (since options are shown in the order
  they appear in the *.el files), or by adding a few :set-after attributes.

** Maybe document the features of libraries missing from the manual (or
565 566 567 568 569
  ancillary manuals, including the Lisp manual in some cases).
  This is not worth doing for all of these packages and we need not
  aim for completeness, but some may be worth documenting.

  Here's a list which is probably not complete/correct: align, allout,
Xue Fuqiao's avatar
Xue Fuqiao committed
  artist, ansi-color, array, calculator, cdl, cmuscheme,
Xue Fuqiao's avatar
Xue Fuqiao committed
  completion, delim-col, dirtrack, double, echistory, elide-head,
  easymenu, expand, flow-ctrl, format [format-alist],
573 574
  generic/generic-x [various modes], kermit, log-edit,
  makesum, midnight [other than in Kill Buffer node],
575 576 577
  mouse-copy [?], mouse-drag, mouse-sel, net-utils, rcompile,
  snmp-mode [?], soundex [should be interactive?], strokes [start from
  the web page], talk, thingatpt [interactive functions?], type-break,
  vcursor, xscheme, zone-mode [?], mlconvert [?], iso-cvt,
579 580
  feedmail [?], uce, gametree, page-ext,
  refbib, refer, scribe, texinfo, underline,
  cmacexp, hideif, mantemp [obsolete?], pcomplete, xml,
582 583
  cvs-status (should be described in PCL-CVS manual); other progmodes,
  probably in separate manual.
Juanma Barranquero's avatar
Juanma Barranquero committed

** Convert the XPM bitmaps to PPM, replace the PBMs with them and scrap
Glenn Morris's avatar
Glenn Morris committed
586 587
  the XPMs so that the color versions work generally.  (Requires care
  with the color used for the transparent regions.)
Dave Love's avatar
Dave Love committed

** Convenient access to the 'values' variable.  It would be nice to have an
Juanma Barranquero's avatar
Juanma Barranquero committed
590 591
  interface that would show you the printed reps of the elements of the
  list in a menu, let you select one of the values, and put it into some
  other variable, without changing the value of 'values'.
Dave Love's avatar
Dave Love committed

** (Controlled by a flag) make open and close syntax match exactly,
  i.e. '(' doesn't match ']'.
Dave Love's avatar
Dave Love committed

597 598
** Specify parameter ID-FORMAT in all calls to 'file-attributes' and
  'directory-files-and-attributes' where attributes UID or GID are used.
599 600 601
  Whenever possible, use value 'string.
  When done, change meaning of default value from 'integer to 'string.
  If value 'integer is used nowhere, remove the parameter ID-FORMAT from
  the definition of 'file-attributes' and 'directory-files-and-attributes'
603 604
  and from the calls.

Kenichi Handa's avatar
Kenichi Handa committed
** Make language-info-alist customizable.  Currently a user can customize
  only the variable 'current-language-environment'.
Kenichi Handa's avatar
Kenichi Handa committed
607 608 609 610 611 612

** Improve language environment handling so that Emacs can fit
  better to a users locale.  Currently Emacs uses utf-8 language
  environment for all utf-8 locales, thus a user in ja_JP.UTF-8 locale
  are also put in utf-8 lang. env.  In such a case, it is
  better to use Japanese lang. env. but prefer utf-8 coding system.
Kenichi Handa's avatar
Kenichi Handa committed

614 615 616 617 618 619 620
** Enhance locale handling:  handle language, territory and charset
  orthogonally and de-emphasize language environments.  Use the locale
  to set up more things, such as fontsets, the default Ispell
  dictionary, diary format, calendar holidays and display, quoting
  characters and phrase boundaries, sentence endings, collation for
  sorting (at least for unicodes), HTTP Accept-language, patterns for
  directory listings and compilation messages, yes-or-no replies,
  common menu items when the toolkit supports it ...  'locale-info'
622 623
  needs extending for LC_COLLATE &c.  [fx started on this.]

Kenichi Handa's avatar
Kenichi Handa committed
624 625 626 627 628 629 630 631 632
** Eliminate the current restriction on header printing by ps-print.
  Currently, a header can contain only single 1-byte charset in
  addition to ASCII.

** In ps-print, provide an user friendly interface to specify fonts.

** Enhance word boundary detection for such a script that doesn't use
  space at word boundary (e.g. Thai).

Kenichi Handa's avatar
Kenichi Handa committed
633 634 635 636
** Implement interface programs with major Japanese conversion server
  in lib-src so that they can be used from the input method
  "japanese".  Currently, most Japanese users are using external
  packages (e.g. tamago, anthy) or an input method via XIM.
Kenichi Handa's avatar
Kenichi Handa committed

Stefan Monnier's avatar
Stefan Monnier committed
** Let LEIM handle the Mode_switch key like XIM does (i.e. a toggle like C-\
Glenn Morris's avatar
Glenn Morris committed
   but which can also be used as a modifier).
Stefan Monnier's avatar
Stefan Monnier committed

Nick Roberts's avatar
Nick Roberts committed
** Improve Help buffers: Change the face of previously visited links (like
Glenn Morris's avatar
Glenn Morris committed
642 643
   Info, but also with regard to namespace), and give the value of
   lisp expressions, e.g auto-mode-alist, the right face.
Nick Roberts's avatar
Nick Roberts committed

** Possibly make 'list-holidays' eval items in the calendar-holidays variable.
Glenn Morris's avatar
Glenn Morris committed
   See thread
Glenn Morris's avatar
Glenn Morris committed
648 649
   [ will look at this after 22.1]

Glenn Morris's avatar
Glenn Morris committed
650 651
** Possibly make cal-dst use the system timezone database directly.
   See thread
Glenn Morris's avatar
Glenn Morris committed

Glenn Morris's avatar
Glenn Morris committed
654 655 656
** Possibly add a "close" button to the modeline.
   The idea is to add an "X" of some kind, that when clicked deletes
   the window associated with that modeline.
Glenn Morris's avatar
Glenn Morris committed

Glenn Morris's avatar
Glenn Morris committed
659 660
* Things to be done for specific packages or features

661 662
** NeXTstep port

663 664 665 666
*** Missing features

This sections contains features found in other official Emacs ports.

**** Support for xwidgets

Emacs 25 has support for xwidgets, a system to include operating
system components into an Emacs buffer.  The components range from
simple buttons to webkit (effectively, a web browser).

Currently, xwidgets works only for the gtk+ framework but it is
674 675
designed to be compatible with multiple Emacs ports.

**** Respect 'frame-inhibit-implied-resize'

When the variable 'frame-inhibit-implied-resize' is non-nil, frames
679 680 681 682 683 684
should not be resized when operations like changing font or toggling
the tool bar is performed.

Unfortunately, the tool bar (and possible other operations) always
resize the frame.

**** Support 'proced' (implement 'process-attributes')
686 687

Unfortunately, a user-level process like Emacs does not have the
privileges to get information about other processes under macOS.
689 690 691 692 693 694 695 696 697 698

There are other ways to do this:

 1) Spawn "ps" and parse the output ("ps" has superuser privileges).

 2) Sign Emacs as part of the distribution process.

 3) Ask the user to self-sign Emacs, if this feature is of interest.

Anders Lindgren <> has implemented
'process-attributes' for macOS, which currently only work when
700 701 702 703 704 705 706 707 708 709 710 711
running Emacs as root.

See this article by Bozhidar Batsov for an overview of Proced:

**** Tooltip properties

Tooltip properties like the background color and font are hard-wired,
even though Emacs allows a user to customize such features.

*** New features

This section contains features unique to Nextstep and/or macOS.
713 714 715

**** PressAndHold for writing accented character

On macOS, many application support the press and hold pattern to
717 718 719 720 721 722 723 724 725 726 727 728
invoke a menu of accented characters.  (See example at .)

Currently, this doesn't work in Emacs.

Note that "ns-win.el" explicitly disables this.

Note: This feature might not be allowed to be implemented until also
implemented in Emacs for a free system.

**** Floating scroll bars

In modern macOS applications, the scroll bar often floats over the
730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762
content, and is invisible unless actually used.  This makes the user
interface less cluttered and more area could be used to contain text.

With floating scroll bars, the user interface would look like it does
when they are disabled today.  However, they will be made visible when
a scroll action is initiated, e.g. by putting two fingers on a

Note: This feature might not be allowed to be implemented until also
implemented in Emacs for a free system.

*** Features from the "mac" port

This section contains features available in the "mac" Emacs port.

As the "mac" port (as of this writing) isn't an official Emacs port,
it might contain features not following the FSF rule "must exist on
free systems".

The "mac" port is based on the Emacs 22 C-based Carbon interface.
It has been maintained in parallel to the official Cocoa-based NS
interface.  The Carbon interface has been enhanced, and a number of the
features of that interface could be implemented NS.

**** Smooth scrolling -- maybe not a good idea

Today, by default, scrolling with a trackpad makes the text move in
steps of five lines.  (Scrolling with SHIFT scrolls one line at a time.)

The "mac" port provides smooth, pixel-based, scrolling.  This is a very
popular features.  However, there are drawbacks to this method: what
happens if only a fraction of a line is visible at the top of a
window, is the partially visible text considered part of the window or
not?  (Technically, what should 'window-start' return.)
764 765 766 767 768 769 770 771 772

An alternative would be to make one-line scrolling the default on NS
(or in Emacs in general).

Note: This feature might not be allowed to be implemented until also
implemented in Emacs for a free system.

**** Mouse gestures

773 774
The "mac" port defines the gestures 'swipe-left/right/up/down',
'magnify-up/down', and 'rotate-left/right'.
775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791

It also binds the magnification commands to change the font
size.  (This should be not be done in a specific interface, instead
Emacs should do this binding globally.)

Note: This feature might not be allowed to be implemented until also
implemented in Emacs for a free system.

**** Synthesize bold fonts

*** Open issues

This section contains issues where there is an ongoing debate.

**** Key bindings of CMD and ALT

Currently in the "ns" port, ALT is bound to Meta and CMD is bound to
Super -- allowing the user to use typical macOS commands like CMD-A to
793 794 795 796 797 798 799 800 801 802 803 804 805 806
mark everything.

Unfortunately, when using an international keyboard, you can't type
normal characters like "(" etc.

There are many alternative key bindings.  One solution is to bind CMD
to Meta and pass ALT to the system.  In fact, this is what Emacs did up
to, and including, version 22.  Also, this is how the "mac" port binds
the keys.

One could envision asymmetrical variants as well, however, this is
inappropriate for the default setting.

See the discussion on emacs-devel:
807 808
809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825

*** Internal development features

**** Regression test system (or at least a checklist)

Today, after each change to the user interface, Emacs must be manually
tested.  Often, small details are overlooked ("Oh, I didn't test
toggling the tool-bar in one of the full screen modes, when multiple
frame were open -- silly me.")

It would be an enormous help if this could be tested automatically.
Many features are generic, however, the NS interface provides a number
of unique features.

**** Existing packages

Note that there is a generic UI test named frame-test.el, see
826 .
827 828 829 830 831 832 833 834 835 836 837 838 839
The NS interface passes this, with the exception of two toolbar-related errors.

**** Anders frame test

Anders Lindgren <> has implemented some (very basic)
tests for full screen, toolbar, and auto-hiding the menu bar.

**** Make sure all build variants work

Emacs can be build in a number of different ways.  For each feature,
consider if is really is "NS" specific, or if it should be applied to
all build versions.

840 841
- With the "NS" interface.  This is the normal way to build Emacs on

- With the "X11" interface.  On macOS, this is mainly of interest to
844 845 846 847 848 849
  developers of Emacs to get a "reference" interface implementations.
  However, it might be of interest for people working remotely, as X11
  applications can be used over a network connection.

- Console only.

850 851
*** Bugs

**** Incorrect translation of Super modifier with Ctrl or Meta on macOS

When pressing 'M-s-a', Emacs replies "M-s-å is undefined".  What
happened is a mix of Emacs view that Meta and Super has been pressed,
856 857
and macOS view that ALT-a should yield "å" (U+00E5 LATIN SMALL LETTER

859 860
The bug reports suggest two different patches; unfortunately, neither
works properly.  For example:
861 862 863 864 865 866 867

   Use a Swedish keyboard layout

   (setq ns-alternate-modifier nil)


Today, this correctly yields that s-] is undefined.  With either
869 870 871 872 873
of the two patches, Emacs responds that s-9 was pressed.

More investigation is needed to fix this problem.

874 875 876
877 878 879 880 881 882 883 884 885 886 887 888 889 890 891

**** Toggling the toolbar in fullheight or maximized modes

The toolbar, in the NS interface, is not considered part of the text
area.  When it is toggled, the Emacs frame change height accordingly.

Unfortunately, this also occurs when the frame is in fullheight or
maximized modes (N.B. this is not the same as "fullscreen").  The
effect is that the full frame size either increases (stretching down
below the lower edge of the screen) or decreases (leaving space
between the lower edge of the frame and the lower edge of the screen).

A better solution would be for the frame to retain its size,
i.e. change the text area.

This is related to the 'frame-inhibit-implied-resize' issue.

**** The event loop does not redraw.
895 896 897 898 899 900
A problem is that redraw don't happen during resize,
because we can't break out from the NSapp loop during resize.
There was a special trick to detect mouse press in the lower right
corner and track mouse movements, but this did not work well, and was
not scalable to the new Lion "resize on every window edge" behavior.
[As of trunk r109635, 2012-08-15, the event loop no longer polls.]

**** (mouse-avoidance-mode 'banish) then minimize Emacs, will pop window back
up on top of all others (probably fixed in bug#17439)
904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922

**** free_frame_resources, face colors

**** Numeric keysetting bug.

*** Mac-related

**** Open file:/// URLs.

**** Put frame autopositioning into C code somewhere -- if loc = same, offset.

**** Automap ctrl-mouse-1 to mouse-3.

**** Deal with Finder aliases somehow.

**** Ctrl-F2 won't pull up menus.

*** Other / Low Priority:

Juanma Barranquero's avatar
Juanma Barranquero committed
**** Better recognition of Unicode scripts / Greek / composition.
924 925 926

**** Undo for color-drag face customization.

927 928
** Bidirectional editing

929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946
*** Support reordering structured text
Two important use cases: (1) comments and strings in program sources,
and (2) text with markup, like HTML or XML.

One idea is to invent a special text property that would instruct the
display engine to reorder only the parts of buffer text covered by
that property.  The display engine will then push its state onto the
iterator stack, restrict the bidi iterator to accessing only the
portion of buffer text covered by the property, reorder the text, then
pop its state from stack and continue as usual.  This will require
minor changes in the bidi_it structure.

This design requires Lisp-level code to put the text properties on the
relevant parts of the buffer text.  That could be done using JIT
fontifications, or as a preliminary processing when the file is
visited.  With HTML/XML, the code that puts text properties needs to
pay attention to the bidi directives embedded in the HTML/XML stream.

947 948 949
*** Allow the user to control the direction of the UI

**** Introduce user option to control direction of mode line.
950 951
One problem is the header line, which is produced by the same routines
as the mode line.  While it makes sense to have the mode-line
952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969
direction controlled by a single global variable, header lines are
buffer-specific, so they need a separate treatment in this regard.

**** User options to control direction of menu bar and tool bar.
For the tool bar, it's relatively easy: set it.paragraph_embedding
in redisplay_tool_bar according to the user variable, and make
f->desired_tool_bar_string multibyte with STRING_SET_MULTIBYTE.  Some
minor changes will be needed to set the right_box_line_p and
left_box_line_p flags correctly for the R2L tool bar.

However, it makes no sense to display the tool bar right to left if
the menu bar cannot be displayed in the same direction.

R2L menu bar is tricky for the same reasons as the mode line.  In
addition, toolkit builds create their menu bars in toolkit-specific
parts of code, bypassing xdisp.c, so those parts need to be enhanced
with toolkit-specific code to display the menu bar right to left.

Glenn Morris's avatar
Glenn Morris committed
970 971 972 973 974 975 976 977 978 979 980
** Custom

*** Extend :set-after to also mean initialize after.
If defcustom A specifies :set-after '(B), then if a user customizes
both A and B, custom will set A after B.  But if the user only customizes
A, then if B is already defined, it gets left at its original setting.
Instead, if B has not been customized it should be re-initialized
(on the assumption that the default value depends on A).
See the places where we manually call custom-reevaluate-setting,
such as for mail-host-address and user-mail-address in startup.el.

Glenn Morris's avatar
Glenn Morris committed
981 982 983 984
** ImageMagick support

*** image-type-header-regexps priorities the jpeg loader over the
ImageMagick one.  This is not wrong, but how should a user go about
Paul Eggert's avatar
Paul Eggert committed
preferring the ImageMagick loader?  The user might like zooming etc in jpegs.
Glenn Morris's avatar
Glenn Morris committed
986 987 988 989

Try (setq image-type-header-regexps nil) for a quick hack to prefer
ImageMagick over the jpg loader.

*** For some reason it's unbearably slow to look at a page in a large
Glenn Morris's avatar
Glenn Morris committed
991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004
image bundle using the :index feature.  The ImageMagick "display"
command is also a bit slow, but nowhere near as slow as the Emacs
code.  It seems ImageMagick tries to unpack every page when loading the
bundle.  This feature is not the primary usecase in Emacs though.

ImageMagick 6.6.2-9 introduced a bugfix for single page djvu load.  It
is now much faster to use the :index feature, but still not very fast.

*** Try to cache the num pages calculation.  It can take a while to
calculate the number of pages, and if you need to do it for each page
view, page-flipping becomes uselessly slow.

*** Integrate with image-dired.

1005 1006
*** Integrate with docview.

Glenn Morris's avatar
Glenn Morris committed
1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017
*** Integrate with image-mode.
Some work has been done, e.g. M-x image-transform-fit-to-height will
fit the image to the height of the Emacs window.

*** Look for optimizations for handling images with low depth.
Currently the code seems to default to 24 bit RGB which is costly for
images with lower bit depth.

*** Decide what to do with some uncommitted imagemagick support
functions for image size etc.

Glenn Morris's avatar
Glenn Morris committed
1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145
** nxml mode

*** High priority

**** Command to insert an element template, including all required
attributes and child elements.  When there's a choice of elements
possible, we could insert a comment, and put an overlay on that
comment that makes it behave like a button with a pop-up menu to
select the appropriate choice.

**** Command to tag a region.  With a schema should complete using legal
tags, but should work without a schema as well.

**** Provide a way to conveniently rename an element.  With a schema should
complete using legal tags, but should work without a schema as well.

*** Outlining

**** Implement C-c C-o C-q.

**** Install pre/post command hook for moving out of invisible section.

**** Put a modify hook on invisible sections that expands them.

**** Integrate dumb folding somehow.

**** An element should be able to be its own heading.

**** Optimize to avoid complete buffer scan on each command.

**** Make it work with HTML-style headings (i.e. level indicated by
name of heading element rather than depth of section nesting).

**** Recognize root element as a section provided it has a title, even
if it doesn't match section-element-name-regex.

**** Support for incremental search automatically making hidden text visible.

**** Allow title to be an attribute.

**** Command that says to recognize the tag at point as a section/heading.

**** Explore better ways to determine when an element is a section
or a heading.

**** rng-next-error needs to either ignore invisible portion or reveal it
(maybe use isearch oriented text properties).

**** Errors within hidden section should be highlighted by underlining the

**** Make indirect buffers work.

**** How should nxml-refresh outline recover from non well-formed tags?

**** Hide tags in title elements?

**** Use overlays instead of text properties for holding outline state?
Necessary for indirect buffers to work?

**** Allow an outline to go in the speedbar.

**** Split up outlining manual section into subsections.

**** More detail in the manual about each outlining command.

**** More menu entries for hiding/showing?

**** Indication of many lines have been hidden?

*** Locating schemas

**** Should rng-validate-mode give the user an opportunity to specify a
schema if there is currently none? Or should it at least give a hint
to the user how to specify a non-vacuous schema?

**** Support for adding new schemas to schema-locating files.
Add documentElement and namespace elements.

**** C-c C-w should be able to report current type id.

**** Implement doctypePublicId.

**** Implement typeIdBase.

**** Implement typeIdProcessingInstruction.

**** Support xml:base.

**** Implement group.

**** Find preferred prefix from schema-locating files.  Get rid of

**** Inserting document element with vacuous schema should complete using
document elements declared in schema locating files, and set schema

**** Add a ruleType attribute to the <include> element?

**** Allow processing instruction in prolog to contain the compact syntax
schema directly.

**** Use RDDL to locate a schema based on the namespace URI.

**** Should not prompt to add redundant association to schema locating file.

**** Command to reload current schema.

*** Schema-sensitive features

**** Should filter dynamic markup possibilities using schema validity, by
adding hook to nxml-mode.

**** Dynamic markup word should (at least optionally) be able to look in
other buffers that are using nxml-mode.

**** Should clicking on Invalid move to next error if already on an error?

**** Take advantage of a:documentation.  Needs change to schema format.

**** Provide feasible validation (as in Jing) toggle.

**** Save the validation state as a property on the error overlay to enable
more detailed diagnosis.

**** Provide an Error Summary buffer showing all the validation errors.

Paul Eggert's avatar
Paul Eggert committed
**** Pop-up menu.  What is useful?  Tag a region (should be grayed out if
Glenn Morris's avatar
Glenn Morris committed
1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347
the region is not balanced).  Suggestions based on error messages.

**** Have configurable list of namespace URIs so that we can provide
namespace URI completion on extension elements or with schema-less documents.

**** Allow validation to handle XInclude.

**** ID/IDREF support.

*** Completion

**** Make it work with icomplete.  Only use a function to complete when
some of the possible names have undeclared namespaces.

**** How should C-return in mixed text work?

**** When there's a vacuous schema, C-return after < will insert the end-tag.
Is this a bug or a feature?

**** After completing start-tag, ensure we don't get unhelpful message
from validation

**** Syntax table for completion.

**** Should complete start-tag name with a space if namespace attributes
are required.

**** When completing start-tag name with no prefix and it doesn't match
should try to infer namespace from local name.

**** Should completion pay attention to characters after point?  If so, how?

**** When completing start-tag name, add required atts if only one required

**** When completing attribute name, add attribute value if only one value
is possible.

**** After attribute-value completion, insert space after close delimiter
if more attributes are required.

**** Complete on enumerated data values in elements.

**** When in context that allows only elements, should get tag
completion without having to type < first.

**** When immediately after start-tag name, and name is valid and not
prefix of any other name, should C-return complete on attribute names?

**** When completing attributes, more consistent to ignore all attributes
after point.

**** Inserting attribute value completions needs to be sensitive to what
delimiter is used so that it quotes the correct character.

**** Complete on encoding-names in XML decl.

**** Complete namespace declarations by searching for all namespaces
mentioned in the schema.

*** Well-formed XML support

**** Deal better with Mule-UCS

**** Deal with UTF-8 BOM when reading.

**** Complete entity names.

**** Provide some support for entity names for MathML.

**** Command to repeat the last tag.

**** Support for changing between character references and characters.