% -*- Mode: Outline; -*- % @(#) 0readme % Last edited: Mon Aug 24 08:03:14 1998 -- Danny T. Quah (dquah@twinpeaks.lse.ac.uk) % $ % Revision History: % % Wed Mar 11 18:32:11 1998 -- Danny Quah (dquah@twinpeaks.lse.ac.uk) % linux 2.0.32 first available; sparc-solaris will have to go away % soon % % Thu Apr 3 18:30:15 1997 -- Danny Quah (dquah@exy.lse.ac.uk) % Made linux and solaris binaries available, on a trial basis % % Wed Mar 19 13:40:32 1997 -- Danny Quah (dquah@exy.lse.ac.uk) % Move from ftp://exx to http://econ. Who doesn't have a Web % browser still? % % Thu Dec 19 19:03:00 1996 -- Danny Quah (dquah@exz) % The DOS/* distribution no longer exists. A hard disk crash took % everything with it, and since I don't [and plan never to] have a % Wintel machine... % % Tue Jul 30 17:42:55 1996 -- Danny Quah (dquah@exz) % * This is the distribution point for Danny Quah's econometrics shell (tsrf). All files require binary transfer: *.tar.gz indicates a gzip'ed (GNU zip) tar archive. *.zip indicates a zip archive. To get tsrf, point your Web browser at (URL) http://econ.lse.ac.uk/~dquah/d/tsrf/ and go from there. My Web homepage also has pointers to papers I've written using tsrf. * {\TSRF} uses a couple of font files for postscript generation. These are plstnd.font and plxtnd.font (although these names are irrelevant to the typical user); on my version of {\TSRF} (the default), these font files are in the directory /usr/local/lib. If the user is on a network (say) or some other installation where he or she is not free to put these font files there, simply put them anywhere else, but then use the environment variable TFPATH to indicate the directory that they are in (e.g., /usr/dquah/c/lib/fonts). Note two things: (1) do not end TFPATH with '/'. (2) The files themselves do not need to be named explicitly, just the directory they are in. * Compiling the shell from source is, I am told, not a trivial matter. I don't know: I only ever type "make" and then "make -f make-skl" when I need to generate a new tsrf. This works for me on every unix machine I use (Linux 1.2.13, Linux 2.0.32, Sparc Solaris 2.x, Next OS 2.x+, ULCC's Fujitsu and Convex machines, etc.). Reconfiguring it for a yet another new unix environment is something I cannot afford to take the time to investigate. If you are comfortable unpacking any of Project GNU's excellent tools under unix, then you're in business, no sweat (or maybe just a little). If you're not, well, it's a very useful little piece of human capital to develop for the weightless economy. * gunzip source should be available on prep.ai.mit.edu. * Remember, all files should be downloaded binary. ** 0t.dvi.gz --- gzip'ed dvi file containing the tsrf manual. After uncompressing with gunzip, this can be viewed online by a dvi previewer (such as TeXView or xdvi). [about 150 pages] ** 0t.ps.gz --- Postscript version of the manual (Type1, formatted to look ok on either A4- or letter-sized paper) ** plplot.tar.gz --- source for plplot (PostScript graphics) ** tsrf[0-2].tar.gz --- C/bison source for tsrf. ** 1perl.zip --- miscellaneous perl and sh scripts. ** tsrf-intel-linux-2.0.32.gz --- gzip'ed Intel linux 2.0.32 binary ** font-intel-linux-2.0.32.tar.gz --- gzip'ed gtar archive of plstnd.font and plxtnd.font for Intel linux 2.0.32 [I can no longer afford to maintain all three different platforms. Soon, intel-linux-1.2.13 and sparc-solaris versions will no longer be available (or at least no longer continue to be updated).] ** tsrf-intel-linux-1.2.13.gz --- gzip'ed Intel linux 1.2.13 binary ** font-intel-linux-1.2.13.tar.gz --- gzip'ed gtar archive of plstnd.font and plxtnd.font for Intel linux 1.2.13 ** tsrf-sparc-solaris.2.5.gz --- gzip'ed Sparc Solaris 2.5 binary ** tsrf-sparc-solaris.2.5.tar.gz --- gzip'ed gtar archive of plstnd.font and plxtnd.font for Sparc Solaris 2.5 Read the installation pages in the manual (0t.dvi) before unpacking. * BUGS ** On a number of operations (graphing for instance), tsrf should check if the entire data series is unavailable. Instead, infinities get added and subtracted, and the program loops forever on the NeXT. On some other machines, an overflow condition is detected, and tsrf crashes. These kinds of boundary conditions don't occur frequently---and are really the user's fault in any case---so that I can't justify my time putting in a workaround. Take care, and have a nice day. ** I know that the internal memory management and linked-list data structures here are a joke. But, hey, they work for me, and are easy to use---they are just inefficient. Moreover, if that's what the bottleneck is in your tsrf code, then you really should be using a different package. Write your own command language, why don't you? % eof 0readme