*  From Lindsay F. Marshall <Lindsay.Marshall%newcastle.ac.uk@munnari.oz>
   The domain name munging (i..e a.b.c -> b.c -> c) doesnt
   seem to work at all, and of course its the wrong way round
   for the uk anyway.

*  From Ian Donaldson <rcodi@melomys.co.rmit.oz>
   From Daniel Klein <dvk@SEI.CMU.EDU>
   Is it possible to have the faces cleared automatically once the
   mail has been read?  I'm clearing by clicking on the window at
   the moment.

*  From Pat Lashley <patl@Sun.COM>
   Have you given any thought to how to support multiple windowing systems at
   execution time?  I am currently using the sun386i version of architecture
   independant /vol trees to make locally installed programs available
   transparently to both sun386i and sun3 users; but this does not address the
   possibility of a single user who sometimes uses SunView, sometimes NeWS,
   sometimes X, etc.. I have considered retaining the names sv_faces, x11_faces,
   etc., with a customized version of start_applic which would determine which
   windowing system is in use, and envoke the appropriate binary. The difficulty
   is in how to determine which system is in use.

*  From: Chris Orgill, cho%cs.aber.ac.uk@uunet.uu.net
   How about allowing the Sunview version to use arbitrary size Sun icons ?

*  From Lindsay F. Marshall <Lindsay.Marshall%newcastle.ac.uk@munnari.oz>
   There is a byte-order assumption somewhere that screws up
   any X bitmaps read in from a file - why dont you use the
   procedure provided by the system rather than writing your
   own?? That copes with the problems correctly.

*  The NeWS version needs to be finished.
   Remaining problems are:

    (a) There are two warnings on compiling news.c, due to FILE operations
        being performed on PSFILE variables.

    (b) The new input functionality (click left/middle mouse, raise window,
        Del, ^L and q) has to be added in.

*  From Pat Lashley <patl@Sun.COM>
   There is no reason that FaceSaver images couldn't be parsed by faces and
   sent to NeWS with the appropriate transformation matrix to squeeze it down
   to icon size.  Let the NeWS PostScript interpreter do the bit
   compression/dithering.

*   From Pat Lashley <patl@Sun.COM>
    An entry of the form `+domainname=community' in the machines.tab file to
    indicate that if a hostname is found without domain specifier, and it
    doesn't match any other machines.tab entry, and YP (oops, NIS) access to
    a server for `domain' is available, and the host matches a key in that
    domain, consider it part of the indicated community. `+=community'
    indicates that the current NIS domain should be used.  This should
    probably be enabled by a compile-time option for those who don't run NIS.

*   From Pat Lashley <patl@Sun.COM>
    An option to display unknown users individually with the username,
    instead of clustered under a single `unknown' user.  Really hot would be
    some way to specify this on a per-community basis...

*   From Pat Lashley <patl@Sun.COM>
    Accept SIGHUP as an indication that various internal databases should be
    updated from the files (i.e. machines.tab), and that the display should be
    immediatly updated to reflect current state instead of waiting for the
    polling period to end.

*   From Pat Lashley <patl@Sun.COM>
    Accept SIGALRM as an indication that the current polling period should be
    immediatly terminated, and the display updated.

*   From Pat Lashley <patl@Sun.COM>
    More internal comments and a general design document.

*   Provide a template of what each of the independent graphics routine do,
    to aid porting to new graphics environments.

*   From John Mackin <john@syd.dit.csiro.au>
    If the mailbox is empty, or contains just one mail item, faces -a starts
    out 10 icon-widths long (with the right-hand part being white instead of
    background grey), then shrinks to one icon-width as soon as a mail item
    arrives or you delete the mail.  I don't care about this since I don't
    use -a, but what I _would_ like to see (here comes the wishlist item) is
    an option that you can use with monitor-new that keeps the window always
    10 (or -c if specified) icon-widths wide, but adds extra rows of icons as
    needed if more than 10/-c new mail items arrive between clearings of the
    window, and shrinks the window back to one row/-c widths when it is
    cleared.

*   From Pat Lashley <patl@Sun.COM>
    Change window header to show what is being monitored (mail, lp1, users,
    etc.)

*   From Pat Lashley <patl@Sun.COM>
    A mixed-mode version which would use xview for the basic windowing and
    property sheets; but use NeWS for the rendering. It is, of course,
    dependant upon having the X/NeWS server, but there may be some way to
    determine at run-time whether the NeWS operations are available, or
    whether to stick to XView/X11 mechanisms.

*   Get the SunView, XView and NeWS drivers to use the -bg and -fg color
    options.

*   From John Mackin <john@syd.dit.csiro.au>
    A comment on the domain lookup algorithm. According to the manual page:
 
           $DIR/machine.dom.ain/uid/iconname
           $DIR/dom.ain/uid/iconname
           $DIR/ain/uid/iconname 
           $DIR/misc./uid/iconname 
           $DIR/machine.dom.ain/unknown/iconname 
           $DIR/dom.ain/unknown/iconname 
           $DIR/ain/unknown/iconname 
           $DIR/misc./unknown/iconname 
  
    This leads to a top-level directory with a _lot_ of subdirectories.
    I think this is responsible for much of faces' CPU time; it's well
    known that large directories are a UNIX performance no-no. What I
    would like to see instead (or in addition, or as an option, perhaps
    compile-time) is the following.

           $DIR/ain/dom/uid/iconname
           $DIR/ain/uid/iconname 
           $DIR/misc./uid/iconname
           $DIR/ain/dom/machine/unknown/iconname
           $DIR/ain/dom/unknown/iconname
           $DIR/ain/unknown/iconname
           $DIR/misc./unknown/iconname

    The idea is that at each level, if the next component exists and is
    a directory, you descend into it and look there.

*   From John Buckle <johnb@cs.su.oz.au>
    With faces v1.4, I've been having trouble in sorting out my faces
    directory path. On the old version (v1.3) I used -f .faces, this meant
    that my directory was searched last which also meant that my misc. face
    was used.  Using FACEPATH I haven't been able to do this, it either seems
    to use one directory or the other, but not both.
    
    	FACEPATH=:.faces	- uses /n/faces/misc. for things stored .faces

    	FACEPATH=.faces:	- uses my misc. for things in /n/face

    Further comment from John Mackin <john@syd.dit.csiro.au>
    This effectively means that under 1.4, you cannot have your own
    misc./unknown icons. The first directory in the FACEPATH to contain a
    misc./unknown icon will be the last directory searched; any directories
    after that will be silently ignored. An unfortunate sude-effect of the
    make_iconname() routine.

*   From Lindsay F. Marshall <Lindsay.Marshall@newcastle.ac.uk>
    With the X11 version under twm, there is a notitle white line problem.
    Also seen under Open Windows. This shows up as a thick white line along
    the bottom of my faces window. The white bar at the bottom is exactly
    twm title sized......

*   From Lindsay F. Marshall <Lindsay.Marshall@newcastle.ac.uk>
    The button1clear bug is weird - you click button1, nothing happens
    until you click button two!!

*   From John Mackin <john@syd.dit.csiro.au>
    If you have a machine.tab entry in an earlier directory in
    FACEPATH instead of a symlink, further directories down the
    FACEPATH are not searched.

*   From Mike Bolotski <misha@ai.mit.edu>
    We've made (and are planning to make more) changes to the faces package.

    Some changes:

    - add flag to allow faces to be compiled to always run -U.

    - added EXCLUDE_UPDATE_DOMAIN so that mail from local users does not
      result in continuous updates

    - fixed icon2ikon to understand the icon format as output by the PBM
      package (no leading 0x)

    - fixed icon2ikon to extract the center 48 pixels instead of just the top
      48.

    Planned changes:

    - change the basic data structure to either a hash table or a balanced
      tree

    - automatic clearing of old faces once the mail has been read (does
      that exist already?)

    - hooks to emacs (RMAIL and VM) to allow automatic X-Face inclusion)

    - lots more, since several students at the AI Lab are fascinated
      by the package.

*   From Ian Darwin <ian@sq.com>
    Need to do a version of faces for tNt v2.0.

*   From Michael Schmidt <michael@uni-paderborn.de>
    With NIS (YP domain fb17i, mail domain uni-paderborn.de) I want to put
    a line "fb17i=uni-paderborn.de" in the machine.tab, so that for all
    hosts in the YP domain faces are looked up under their mail domain.
    [Set of diffs included].

*   From Lindsay F. Marshall <Lindsay.Marshall@newcastle.ac.uk>
    Some suggestions:

    - people.tab/machine.tab per face database

    - caching of icons to reduce file accesses

    - re-reading people.tab/machine.tab when modified

    - access to gnu finger (face) service

*   Add an examples configuration section to the README file for a variety
    of machines.

*   From John Mackin <john@ditsydh.syd.dit.csiro.au>
    Support color XPM image format.

*   Compiling the NeWS version gives an undefined symbol at link time for
    _draw_screen.

*   Convert ~/.facerc to Xdefaults format. Add in support for the SunView and
    NeWS versions to be able to read this file.

*   Possibly convert command line arguments to word equivalents.

*   Convert all fixed length string handling in faces to use malloced strings.

*   From Steve Kinzler <kinzler@iuvax.cs.indiana.edu>
    A major inadequacy with the current -e mechanism is the inability (or
    inconvenience) of faces application programs/scripts to save any state
    -- they're re-invoked fresh every time period. We may need application
    programs that run concurrently with faces with bi-directional
    communication.

*   From Mark Shand <snand@prl.dec.com>
    Add a one or two word hashed key of the X-Face: bitmap to each internal
    faces record, so that two identical X-Face'd records from the same user
    can just generate one icon with the -a option.

*   From Daniel Klein <dvk@SEI.CMU.EDU>
    It would be nice if I could display faces vertically and not just
    horizontally, (gee, a scroll bar would be great, but not essential).

*   From Daniel Klein <dvk@SEI.CMU.EDU>
    Autoscaling the faces window would be nice too (so I don't have empty
    squares)

*   From John Mackin <john@syd.dit.csiro.au>
    A huge performance win for people just running faces, monitoring for new
    mail, can be had in one of two ways:

    (1) read and scan the mail file backwards just extract the required
        number of messages (10 by default).
    (2) scan all the messages, but just cache the info for locating the face
        for the last -c number of messages, and defer looking up icons until
        you hit EOF.

    The first suggestion is harder to program neatly and efficiently, but
    much cuter; it does, of course, defeat the kernel's read-ahead, but what
    the hell.

*   From John Mackin <john@syd.dit.csiro.au>
    As it currently stands, it's not possible to use machine.tab and
    people.tab, except in the last directory tree on the faces search path.

*   From Victor A. Riley <var@iris.brown.edu>
    For X11, you copy faces to xfaces yet do nothing after that.  You should
    at least have an install.x11 (and install.xview, install.news, ...)
    What about the man page for xfaces?  I know it is the same as the faces
    man page, but you should probably link the two man pages together.

*   From Mark Shand <shand@prl.dec.com>
    Uncompface is still far from robust in the face of maliciously or
    erroneously constructed faces.  I just received an X-Face: header in
    which the X-Face: keyword was duplicated.  Not only did this face
    decode incorrectly, but subsequent, correctly formed face headers
    were also corrupted.  Uncompface() needs to have bullet-proofing added.

*   From Victor A. Riley <var@iris.brown.edu>
    Modify the Makefile.dist file so that the 'make tables' entry installs
    the misc. faces in the correct place.

*   From Steve Kinzler <kinzler@iuvax.cs.indiana.edu>
    I wonder, since we seem to be using xbm format as the standard now,
    should it be made the default format?  That is, a HOST/USER file would
    be assumed to be xbm instead of ikon, in contrast to HOST/USER/face.xbm.
    Just something to consider ...

*   From Lindsay F. Marshall <Lindsay.Marshall@newcastle.ac.uk>
    I still want faces to work the way that the old V8 vismon did which is
    that new faces slid gently in from the side of the window and didn't
    just pop up suddenly.

*   From Philippe-Andre Prindeville <philipp@inf.enst.fr>
    Another suggestion about face_update:  it would be nice if it did
    a "umask 022" at the very beginning, and if it were explicitly
    installed to run as "daemon,daemon" as setuid/setgid.  We run our
    mailer in locally in "secure" mode, God knows why, and this
    causes ownership and security problems...

*   From Alan Skea <skea@vast.eecs.unsw.oz.au>
    Sometime between versions 1.4.9 and 1.5.1, some change to
    compface/uncompface has made it produce different output. The face
    resulting from the uncompression of a 1.4.9 X-Face: using the 1.5.1
    algorithm seems to shift the whole image to the left a bit (or maybe
    two - I don't really have time to check this out thoroughly) and
    introduces some corruption into the bottom right of the face.  Caveat:
    these observations are made on the basis of a sample space of 1 face.

    From John Mackin <john@syd.dit.csiro.au>
    The situation, so that everyone knows (Rich already knows) is this:
    there are definite problems with post-1.4 compression and decompression.
    I am liasing with James Ashton, but he is very busy too now and has
    little time to look at it; what I was hoping was that I could work out
    what was wrong and then discuss the fix with him.

*   From Dirk Craeynest <dirk@cs.kuleuven.ac.be>
    With the remote hostname/user monitoring facility within faces it
    would  be nice to have an option specifying that every user should
    only be shown once (perhaps with the number of logins).

*   From Daniel Glazman <glazman@inf.enst.fr>
    There is currently a byte-order assumption in the way that Sun icon, and
    Blit ikon files are read.

*   Convert the [un]compface code to read/write X bitmap files (as opposed to
    Blit ikon files).

*   From Rohit Aggarwal <Rohit.Aggarwal@Corp>

    - It would be nice to change the -v option such that the version is
      displayed in the window-tool header instead of printing it out on
      the screen.

    - Add:  tbl faces.man | {nroff,troff} -man   to the README to print
      out the man page.

    - It would be nice to a default.au file that is played if a X.au
      file is not found in the user directory.  This is similar to the
      default icon functionality you have.  This way one can have a common
      X.au file for each community.  You can have a directory called audio
      in which one can have the default sound file.

    - I have just updated to OW3.0pre-fcs and in the faces window where the
      user-name does not fill the entire icon-bits the workspace color is
      showing through.  This did not happen in OW2.0.  Not a big deal just
      thought I might let you know.

*   Integrate the compface and xface directories into one set of sources that
    do both jobs.

*   From Jim R. Oldroyd <jr@sequoia.com>
    An option to get faces to read a single email message containing the
    X-Face on the stdin, and to produce the face-update email, containing the
    48x48x1 on the stdout. (I.e., like -U, but output not mailed.)
    This will allow users to install a separate update scheme, even if the
    administration won't allow a `facemaker' alias!

*   process_info() in mon.c needs to be reworked so that make_iconname()
    is not called if this is an X-Face:, and audio support isn't enabled
    and there is no updating of the faces database (-U command line option).

*   Faces no longer chdir's to the top of each facepath element before it does
    it's searching. This functionality (somehow) needs to be reinstated.

    Comment from Steve Kinzler:
    To fix it within faces, I could see implementing the concept of a lookup
    root directory.  Just during face lookup, the program would chdir here
    and then any relatively specified (ie, don't begin with "/") db
    directories would be relative to this directory.  The program already
    has a default facepath=/usr/local/faces -- we could make this the
    default lookup root directory with the default facepath=".".

*   From Harald Vogt <harald@cs.ruu.nl>
    Faces shouldn't use an obsolete format for the X-Face: image. Something
    like XBM would be much more appropriate.

    (Why should I have to install PBMplus in order to convert my XBM image to
     Blit ikon format?)

*   Optimise.
