%% /usr/local/src/mg/README.UTAH, Thu May 19 16:55:43 1994 %% Edit by Nelson H. F. Beebe Distribution notes for the mg full-text retrieval system Nelson H. F. Beebe Center for Scientific Computing Department of Mathematics University of Utah Salt Lake City, UT 84112 USA Email: beebe@math.utah.edu (Internet) [19-May-1994] This distribution of the mg full-text retrieval software is intended only for users at the University of Utah, and only for a limited period until the improvements communicated back to the original authors have been incorporated into their master distribution. The book reference is: @String{pub-VNR = "Van Nostrand Reinhold"} @String{pub-VNR:adr = "New York, NY, USA"} @Book{Witten:MG94, author = "Ian H. Witten and Alistair Moffat and Timothy C. Bell", title = "Managing Gigabytes: Compressing and Indexing Documents and Images", publisher = pub-VNR, year = "1994", ISBN = "0-442-01863-0", LCCN = "TA1637 .W58 1994", address = pub-VNR:adr, pages = "xiv + 429", price = "US\$54.95", acknowledgement = ack-nhfb, note = "The software for full-text indexing described in this book, and errata for the book, are available for anonymous ftp from \path=munnari.oz.au= in the directory \path=/pub/mg=.", bibdate = "Thu May 12 08:02:54 1994", } There are NO source code changes to this distribution, but there are very extensive revisions to the manual pages, and a small change to the top-level Makefile. The man/man1/README file describes the changes to the man pages. All changes are under RCS control, so you can see exactly what has been changed by doing rcsdiff -r1.1 filename In addition, there is a new top-level subdirectory named utah that contains a Makefile and a revised mg_get script with bibfiles and texfiles targets. This will soon be extended to handle other types of files. I intend to try to get this software to build on all major vendor UNIX platforms (DEC (ULTRIX 4.x + OSF/1 1.x and 2.x), IBM (AIX 3.2.x), HP (HP-UX 9.x), SGI (IRIX 4.x and 5.x), and Sun (SunOS 4.x and Solaris 2.x). It already installs without problems on Sun and MIPS systems, and HP support is partially there. If you do achieve a port to another architecture, please send any changes back to me so I can coordinate them with the developers in Australia. I'm in regular e-mail contact with them. There is no install target in the Makefiles yet: after a successful build, just do something like this: strip bin/* cp bin/* /usr/local/bin cp man/man1/*.1 /usr/local/man/man1 To use xmg, the X Window System interface to mg, you need to install tcl; I have not yet done so. To make use of the utah directory, after installing mg on your system, just copy the two files there (Makefile and mg_get) into a new empty directory named $HOME/mgdata, cd there, and type "make". I find it useful to have this automated by creating a cron entry something like this: 00 04 * * * rsh somehost.utah.edu 'cd $HOME/mgdata; (pwd ; date ; df . ; make ; date ) >>&make.log' This assumes that the login shell is csh or tcsh; otherwise, a cron entry like this will do: 00 04 * * * cd $HOME/mgdata; (pwd ; date ; df . ; make ; date ) >make.log 2>&1 This runs mg every night at 4am; to run it hourly from 7am to 6pm, change the 04 to 07-18. The first time mg is run, it may take 10 to 15 minutes for several hundred files, or perhaps an hour for a few thousand files. Subsequent runs seem to be much faster; for one of our initial users with 1600 files, update runs take less than 2 minutes, so hourly updates are not a problem. The major portion of the time is spent in scanning the file tree with find(1); in my own personal file tree (several hundred MB), the whole process takes 70 minutes. Please report any problems with this distribution back to me, rather than the original authors, so that we can channel communications through a single coordinating point, so as not to swamp the authors who have given us 35K lines of very useful new code, and who are very busy with the start of their Fall Semester (remember, Melbourne is upside-down from us).