1) make sendfax recognize "-h <file>" option to set page headers
  started 20.3.93, GD
  new idea: send "AT+FDT", then (optional) page header, then page data.
  (does not work)
  new idea: customized pbmtog3, without EOP.
  "first cut" done, 21.9.93, use "g3cat"
  Suggestion: use "-h <program>", g3cat output of that program with the
  data page (-> headers with real send time)

  see 119

2) make external configuration possible
   mostly done (1.1.95)

3) #include <config file>
   bad idea, not really useful

6) faxmail - gatway program for mail -> fax
   (started)

7) when (before) receiving a fax, check if there's enough space left
  (amount configurably) -> clewis
  generic "get spool file name", check multiple directories
  (multiple dirs done, code is ugly, clean up)

8) function for creating spool file names (->dir, prefix; <- name, fd)
  with (optional) sequence numbers

14) make mgetty (and g3cat) add an digifax header to incoming faxes (optionally).

16) tiffg3 -> raw g3 / digifax converter

17) library functions for opening / reading g3 files

18) Makefile.<system> / AutoConf?

20) tio - increase abstraction layer even more, make TIO an opaque pointer, so
        that the upper layers do not have to know *anything* about the data
        structure used by tio.c/.h
	(with old/new settings, to avoid surplus system calls)

21) additional flag: skip modem initialization and all, go directly to
    reading of login name (for callback)  (would also help Hylafax)

24) fax_notify_mail: include "realname" (from fax phone book) in Subject:

31) VSI*FAX

34) ungetty mechanism

35) utmp writing / process arguments on SunOS

45) make fax poll server more flexible

51) faxspool.c (sgid fax) -> $LIBDIR/faxspool.sh
    -> on TODO since at least June 1994
    -> 18.11.2002: *done* via "faxq-helper"

51a) have it start faxrunq immediately (configurably)

55) vhangup()? (->Linux)
    (security issues)

57) SIGUSR1 *after* answering -> hangup

59) mgetty/faxrec(): valiate incoming fax data, request retransmission
    partially done, if +FP(T)S:... returns valid values.

60) faxspool - multiple targets in one fax
    call: faxspool tgt file(s)
	  faxspool -m tgt1 tgt2 tgt3 -- file(s)
	  faxspool -M tgt-file file(s)

62) session parameter mismatch (width / length / resolution) -> correct
"on-the-fly" (integrate "smart" g3cat into sendfax.c). If the page is too
long, cut after 2050 (letter) or 2156 (A4) lines.

63) ZyXEL callback (description, "hack" in mgetty.c - ZYXEL_CALLBACK)
    (integrate in state machine)

63b) real callback

66) faxcleanup

73) locks.c: only "guess" that it's a binary lock if it doesn't consist
    entirely of " 0-9\n".

75) dialcode translations in faxspool

90) support ct 

91) g3split

96) number of RINGs changeable without /etc/inittab change
    (done). Changeable without restart?

97) BREAK -> mgetty exits (does it?)

98) fax_wait_for: accept "<junk> +FHNG:..."

107) create own fax file header, with sender ID (if known), resolution,
     and *magic byte*!

108) faxreq (re-queue, remove Status junk from JOB)
     done with "faxq -r" - really delete "Status" lines?

112) sendfax: return +FHNG in exit() code

114) callback: other lines (if available), random delay

116) use more detailed access privileges in faxspool (long distance, local)

118) modem auto-detect not only for voice but for modem type (e.g.
     MultiTech, ZyXEL, ...) in general
     12.12.97: basic work done

119) dynamically generate page headers in faxrunq
     12.12.97: integrate with class 1 work - we have to "g3cat" the page
               data anyway (for padding)

120) with 119, multicasting is easy

121) check disk space before spooling faxes

123) -DNO_MODEM?

124) if write/read in fax_send_page() fails, exit loop (kkeyte)

125) cleanup script that will rollover logs

126) if sendfax is killed, faxrunq assumes "ok" for some strange reason.
     (caused by the way exec() / wait() handle return code / signals)
     10.12.97: resolved for faxrunqd.  Still open for faxrunq???

127) step-by-step installation guide

128) g3topbm: use same decoder as g3view (speed/space tradeoff?)

129) g3topbm: speed up decoding by reading the file as a whole

135) for modems where AT+FAA=1 doesn't work: use SIGUSR2 to toggle
     fax/data (for people knowing what type of call comes in)

136) keep modem in data mode, switch to fax mode (+FAA=1 or 0, depending
     on -D/-F) if desired before sending ATA. Reason: dialout processes
     can safely assume "+FCLASS=0" with this setup.
136a) add signal to switch reception mode for dumb no-adaptive-answering
     modems, see 135)

137) call "new_fax" program for each page during reception (for those
     people that have a fast machine and a slow printer and want to see
     the results immediately. Hi Harald!)

139) RING during initialization -> wait for it to stop, init modem again

141) make lock/log/spool paths configurable

144) if first line in logfile is written *after* setuid() in login.c,
     no opening of LOG_PATH is possible. open log sooner? mode 666?

145) include Device/Hostname in "notify mail"
     [Device done. Do we really want the hostname?]

79) g3cat: fix line width (that is, "repair" broken T.4 files on-the-fly)
147) make "g3cat" fix line width!
     (done, but doesn't seem to work for all receivers) 

151) SCHED option?

152) get bit order for class 2 / class 2 ok / 2.0 for receiving and polling right

153) testing: add some signal that will cause a fax reception to "fail" (RTN)

154) add some way to change runtime configuration

155) fax receiver: ignore signals (USR1, USR2) more completely, only
     barf about "interrupted system call" if SIGARLM

156) if login name contains spaces, split into several arguments
     (*optionally*). Use : "login username ENV1=foo ENV2=bar".

157) add a "what to try if modem init fails" routine (try 19200 bps,
     try sending <DLE><ETX> to get modem back from voice mode, ...)
     (Julian Assange, proff@suburbia.net; ard@cstmel.hobby.nl)

     09.07.96: <DLE><ETX> done, +++ATH moved to that part as well. Baud
     rate switching still open.

160) make "dialin.config" path configurable (and all other paths
     except "login.config" as well). [not sure whether I want this]

161) make path to "mgetty.config" configurable via command line switch

163) SIGUSR2: harden all "byte-receive" functions against EINTR

168) accept error messages while sending a page

169) "faxspool -t" must be smarter: "send at 2:00 tonight" isn't possible now

170) make "waiting for lock file to disappear" smarter: wait for *PID* to
     disappear *or* mtime change [->no need to read() the file every time
     if mtime didn't change and PID still exists] (-f =remco)

171) callback: password!

174) on Dr.Neuhaus: stay in +fclass=2 mode, don't go to +fclass=0
     same for the Hayes Optima (+FAA=1 doesn't work in +fclass=0 mode)
     
175) man page documentation for "login.config" 

177) AIX, native SCO: DTR toggle needs re-open() on device

178) Linux: #ifdef PARANOID -> lock HUPCL (and CLOCAL?) flag in termios
     and do vhangup() stuff

180) switch from policy.h-dist to policy.h with "reasonable defaults"
240) do away with policy.h, make everything run-time configurable

184) make sendfax choose fax device in a "random" fashion (avoid 
     problems when the first modem in a "hunt group" dies (NO DIALTONE))
     (if faxrunqd is used, it will take care of this)

186) make "distinguished ring" run-time-option:
	dist-ring "RING 1" fax
	dist-ring "RING B" fax/data
	dist-ring "RING 3" voice
     (dynamically growing list, appended to "action" list)
	
187) clean up tio handling / gettydefs stuff in mgetty.c

188) init / DCD drop / I/O error: can we catch it with CLOCAL?

189) sendfax: log "hint" if "ignore-carrier" is not set but "should be"

191) watchit.pl: make "remail" smarter

193) makelock: check for existance of lock file *directory*, complain
     (loudly, L_FATAL!) if it doesn't exist.

194) faxspool: add switch to suppress e-mail

195) make "logging *only* to syslog" conditional compile

197) +FAP (Sub-Adressing frames) support?

199) selectively reject fax reception (on sender ID)?
     http://www.vix.com/hylafax/setup-advanced.html#QualifyTSI

200) build kind of "libiberty" with auxiliary functions [split up goodies.c]

201) make "open() on TCP/IP" socket possible in sendfax (for things like
     portmasters, that have special IP ports to access their modem cards)

203) fix ATD/ERROR problem with USR vs. ERROR response in fax_send_ppm

205) vgetty on AIX: make sure VTIME is not set != 0 while waiting with
     select() or poll() -- drivers are dumb and use VTIME for timeout, 
     not the value passed to select()... *sigh*

206) if "Dr.Neuhaus SMARTY" is detected, set some flag to prevent
     switching back to +fclass=0 ("modem_quirks |= NEED_2"?)

207) if "Telejet SaferFAX" set "cls2ok" ("modem_quirks |= PROPER_BOR");

209) faxrm - fix "call with no arguments"

210) fix cnd.c / mgetty.c: clearing of CallerID if call was rejected
     (Frank Piwarski, fpiwarsk@MyMy.com)

211) fix CLOCAL handling in mgetty (->Ted T'so)

212) faxrm: generic "removeit" subroutine, not multiple "rmdir $jobid"

213) add support for Mark Himsley (mark@mdsh.com)'s weird 2400/fax modem,
     requiring multiple baud rate switches

215) implement e-mail to fax gateway using fax addressing as in IETF
     draft "draft-ietf-fax-addressing-02.txt" (or similar)

218) check whether "ATI1" always returns version number info on the
     *first* line on ZyXEL 2864I models, instead of "checksum first"
     (like all other models) --> faxlib.c, "case 28642:"

222) fix "g3cat"'s line width correction algorithm - include black dot.

224) faxspool: permit "*" as part of the phone number -->
     faxrunq(d): make sure that "*" in $phone is never expanded as wildcard!

225) use a non-root user-id (nobody?) for "cnd-program"
225a) change "system" to "build a proper argv[], then fork/setuid/exec/wait"

226) call cnd-program after "min-rings" RINGs, and before waiting for
     the remaining number of RINGs.
     (Also, that way vgetty can control the number of RINGs depending on
     the distinctive RING number / ISDN MSN)
226a) have the cnd-program decide whether to "answer NOW", or "ask me
     again later".

     2018-03-05: sort-of done (226+226a) with cid-program / return-value
     (cid-program can set rings_wanted to "1" [now!] or "some high number",
     after which cnd-program will be run - still not fully flexible, but
     maybe that's not a real use-case anyway)

227) mgetty: add config option to call external program in the case
     of "L_AUDIT / failed" exit's?  Might be helpful for an external
     management system... (Robert Canary)

227a) external scoreboard (shared memory, etc) too keep track of things?

228) find a way to limit the number of "faxrunq"'s that can run in
     parallel (on systems with a very large fax queue, it might happen
     that faxrunq is ran often enough to overflow file/process tables,
     causing everything to break) - Graham Leggett

     WORKAROUND: use faxrunqd if you have many modems and high fax througput.

229) weird idea: use NSFs to communicate version number
     "proceeding", mgetty's own NSF vendor code assigned 2006/06/07

230) add "frontends" section to the documentation (mgetty.texi)

231) pass dist_ring and called_id as env variables to login/new_fax

232) have cndfind() signal back to ring.c if the string found was a
     "caller ID" (for speedups, Caller ID is counted as "RING", to
     avoid waiting for the next real RING which might delay answering 
     by a number of seconds)

233) have different login: prompts for callback/non-callback logins

234) make new_fax path configurable

235) build "faxspool-d" daemon that listens to network requests, plus
     "r-faxspool" client.

236) make Windows client use faxspool-d, not lpd

237) use r-faxspool/faxspool-d locally as well, to avoid world-writeable
     /var/spool/fax directory

238) show vgetty version in "vgetty -V", not only mgetty version

239) make utmp entry (CallerId/Connect) configurable from mgetty.config

241) cnd-program rework (more return codes, call after first RING)

242) document callback and login.conf v2

243) fix page lengths while sending (integrate in "send G3 file with
     byte padding" that's necessary for class 1 anyway)

244) faxrunqd / write combining: if some pages of a "combined" job
     (everything but the first) have already been renamed to f<n>.g3.done,
     sendfax will refuse cooperation ("command line error").  BUG.

245) open anon CVS for mgetty

248) include script to rotate/move log files on a regular basis
     (base on contrib/mvlog, move to tools/, run through sedscript for
     location and naming for the log files)

250) move prototype for time() to mgetty.h, remove everywhere else

252) implement backend code for mgetty/sendfax.config "*-flow" directives

253) rewrite modem ID code (base on ATI9, fall back to ATI if ATI9 
     doesn't work)

254) in vgetty, "voice_log_level" overrides "-x<n>" - cli args should have
     precedence!  Tricky problem (voice/libutil/config.c).

255) if faxrunq(d) can't read config file due to permission problems, don't
     ignore error but print some informative message

256) add easy way to create/remove $FAX_SPOOL_OUT/stop (fax-helper?)
     done: faxq stop / faxq start

257) generalize calling $(MAKE) in subdirs - put all $(MAKE) arguments
     in one common variable

261) faxrunqd: don't count "modem locked" as "failure" for a given job
     (but *do* delay by a certain time)
     (hmmm - maybe not.  If all modems are locked for a long period of
     time, it might be useful to flag the fax as "failed" to get people to
     notice.  Combine with 259 - should resolve problem)

262) faxrunqd: smarter algorithm for "modem is broken"
     (locked = doesn't count, can't init = +1, NO CARRIER = +0.2)

263) faxspool.rules: add sample mappings for PDF (via acroread, xpdf,
     recent ghostscript versions)

264) faxrunqd: timeout on "hanging" sendfax processes
     (timeout > 15 min + 5 minutes * number of pages, or something like that)

265) faxrunqd: write tty to acct.log (format change / WEB gui)

260) faxrunqd: bump up priority of "exit 12" jobs
266) faxunrqd: if a job fails half-way through (12), increase priority
     to re-send missing parts quickly

267) faxrunqd: sort per-modem statistics numerically (watchout: "total")

268) faxrunq+faxrunqd: have configurable counters for 
        "send-a-warning after <x> tries, fail after <y>"
     have separate counter for "modem-locked"
              costing-failures    tries-except-modem-locked    total-tries
       warning        -                     10                    10
       failure        5                     30                   100

269) faxrunqd: have configurable threshold for "everything is broken"
     -> if that threshold is reached, generate STOP file, warning e-mail, 
     and suspend itself  (like "the local in-house PABX is broken")

270) Makefile: have all paths depend on $(root)/  (->Debian)
     use $(root)/ for "make bindist"

271) rethink whole concept of "blocking yes"
	- when is this supposed to be used ("getty" replacement)
	- how is it supposed to work
     and then fix logging in mgetty.c ("waiting..." is not appropriate)

272) have a configurable delay after each "sendfax" call in faxrunq(d)
     (reason: long but not timing-critical out queue, don't starve
     incoming calls. Wouter van Marle, wouter@squirrel-systems.com)

273) update "ZyXEL" modem section in mgetty.texi-in - it's mainly specific
     to the 1496s.  More recent ZyXELs need mentioning.
     ksadil@bigpond.com reports that the U90e works well with vgetty.

274) g32sff  (based on g3cat routines)

275) sff2g3: duplicate previous line on 0xff/0x00 tag

276) fix "make beta" process (no gcc on target machine anymore)
     [26.11.05: moved to new target machine]
276a) add freshmeat net reminder

277) mg_m_init.c/mg_get_ctty() - if setsid() fails, and getppid() != 1,
     mgetty might be run from the command line.  Should we try fork()ing,
     then setsid()'ing, and then getting a new controlling tty?  
     (Things are likely to fail anyway due to missing utmp slots...)

280a) implement BREAK and dual-tilde sending in tools/microcom.c

282) add test code for ring.c, faxlib.c, and other low-level modules
     (overload mdm_read_byte() -> feed various RING strings to ring.c etc)

     t_ring.c: test "action" strings (CONNECT, ERROR, etc.)

283) mvlog: gzip

284) test function for faxq-helper?

286) faxrunqd: add hard (long) timeout to "wait for child processes" so
     that faxrunqd can recover in case something seriously gets stuck

287) run "cvs tag" automatically after "make beta"

288) change all #idents to contain @(#) prefix

289) re-do all the bitmap conversion stuff in faxspool
     (with recent netpbm versions, pgmtopbm doesn't properly work
     anymore, instead you need to use pamditherbw | pamtopnm)

290) (re-)organize fax class 1 "state" - e.g. always use fax_param_d

291) implement T.37 on-ramp and off-ramp gateways

292) implement T.38 on-ramp and off-ramp gateways

293) implement -o 'myoption arguments' flags to mgetty and sendfax, get
     rid of all these flag letters