Notes for websterd version 1.2
Steve Hayman
sahayman@cs.indiana.edu
1991 05 13

This is version 1.2 of my NeXT webster daemon for NeXTOS 2.0.
(The first version of this program was supposedly called version 
2.0; sorry for the confusion, I'm starting to use RCS to log the
changes and it's simpler to call this one 1.2. )
This daemon runs on a NeXT and allows remote (non-Next) machines
on your network to look things up in the NeXT's dictionary, using
either the "webster" or "xwebster" (for X11R4) client programs.

Version 1.2 adds support for the thesaurus, which wasn't present
in version 2.0.0 .   Be sure to read the section of the Installation Notes
(below) about what you have to do to get the thesaurus to work.

The dictionary format changed drastically when NeXT went to 2.0,
so my old version of websterd for NeXT 1.0 doesn't work at all.
Unfortunately to compile this one, you need to ask NeXT to send
you the "libtext" and "libbtree" libraries, since they aren't
included with NeXT OS.  Send a note to "ask_next@next.com", tell
them you would like to get libbtree and libtext.  If you can set
things up so that they can ftp the libraries to you (they're pretty
big) things will go nice and quickly, otherwise you have to send
them floppies or an optical.

I wish I could give you a copy of libbtree and libtext, but I checked
with NeXT and they're not keen on that.


Installation Notes:

1) Get the libbtree and libtext packages from NeXT as described above.
   These will come to you as two large compressed tar files.
   Uncompress and untar them. 
   
   Now you need to put libbtext.a, libbtree.a and the include
   files somewhere where the Makefile can find them.
   Here's one possibility.
    Copy LIBTEXT/libtext.a to /usr/local/lib/libtext.a
    Copy LIBBTREE/libbtree.a to /usr/local/lib/libbtree.a

    mkdir /usr/local/include /usr/local/include/text /usr/local/include/btree

    copy LIBTEXT/include/*.h to /usr/local/include/text
    copy BTREE/usr/local/include/*.h to /usr/local/include/btree

2)  Add this line to /etc/services

	webster		2627/tcp
    
    If, as is likely, you are running "netinfo" you will need to do
    something like "niload services . </etc/services" after you
    update /etc/services.  If you're running Yellow Pages you might
    need to do something similar but yellow-pagish.  

3)  Add this line to /etc/inetd.conf  (this is one line)

    webster		stream	tcp	nowait	nobody	/usr/local/etc/websterd websterd -l -n 129.79.

    The "-l" flag says "please syslog various useful things".  I use that
    for debugging, you might not want it.  Websterd logs things
    as "daemon.info" - if you want this kind of logging, you should
    add a line like this to /etc/syslog.conf and send a HUP to syslogd:

	daemon.info	/dev/console
	    (Use command-C in the File Viewer to bring up a console window)

    or perhaps

	daemon.info	/usr/adm/messages


    The "-n 129.79." says "restrict connections to subnet 129.79.", which
    is our domain here at Indiana.  Replace this by something suitable.
    You can list several domains, like this

	-n 129.79.254.,129.79.253.,129.79.251.

    if you want.

    Note: you *have* to use a -n option.  If you don't, access will
    be restricted to "127.0.0.1" (localhost) only.


4)  Compile and install websterd.  The Makefile right now puts the
    program in /usr/local/etc/websterd - if you don't like this location,
    change the Makefile and the appropriate line in /etc/inetd.conf .

5)  Send a HUP signal to "inetd".  i.e. "kill -HUP <proccess id of inetd>"

6)  Fix the thesaurus.  Run "make repair".

    Why is this necessary?
    Here's the deal.  The NeXT libbtext.a library routines will crash
    your program if you try to use them to read the thesaurus.   I've
    figured out why this is.  Briefly, the thesaurus and dictionary
    files are implemented as BTree files named
    /NextLibrary/References/Webster-Dictionary/Dictionary.nxbf and
    /NextLibrary/References/Webster-Thesaurus/Thesaurus.nxbf.

    Any Btree file can contain one or more indices.  The Dictionary
    file contains both a "headword" index and a "Full Text" index.
    The thesaurus contains only a "headword" index.  (The difference
    is that a headword index records only the "primary" entry for
    any particular word; a Full Text index has an entry for each
    word every time it appears in the dictionary.  So, in the
    dictionary headword index, there is only one entry for the word
    "bird" and if you look up "bird" using that index you'll only
    see the main definition for "bird".  If instead you use the
    "Full Text" index, you'll see the definitions for aepyornis,
    aeire, air sac, albatross, ... and any other word whose definition
    includes the word "bird" somewhere in it.)

    Now, the problem is that the libtext routines cannot deal with
    a BTree file that contains only a Headword index.  I don't know
    why this is, but the referenceOpen() routine will cause a core
    dump if you try to open the thesaurus, since it has only a
    Headword index.  So, I have include a program with websterd
    called "addfulltextindex" which simply creates an empty
    index called "FullTextIndex" in a BTree file.  If you run
    "make repair", this will do 

      addfulltextindex /NextLibrary/References/Webster-Thesaurus/Thesaurus.nxbf
    
    This modifies the thesaurus BTree file slightly, so that it now
    has both a Headword index and FullText index (even though the
    latter index is empty.)   Now the libtext routines will work.

    You have to run 'make repair' once before your websterd can
    access the thesaurus definitions.  websterd checks to see if
    the thesaurus has a Full Text index, and if it doesn't, it produces
    a message saying "Hey, go bug your system administrator".
    


Legal Note:

I have been told informally by a few people at NeXT that using this
dictionary to export webster data to a non-NeXT machine is "illegal".
I find this hard to believe and have not been able to find anything
in the software license agreement that forbids this.  What good is
a dictionary if you can't look up words in it?  It seems to me that
exporting the dictionary data to other non-NeXT machines *belonging
to the same "department"* is a reasonable use of the dictionary.
Please don't export this dictionary to people who wouldn't otherwise
have access to your NeXT.  That's not fair.




Anyway, enjoy.
I welcome any and all feedback.
..Steve
sahayman@cs.indiana.edu