Name: skT88420 Date: 02/17/2000


java version "1.2.2"
Classic VM (build 1.2.2-L, green threads, nojit)


1) HTMLConverter -help text doesn't reflect reality:

   NOTE: On UNIX, each filespec must be qouted. ("*.html" "*.htm")

   Experiments show that this is not true, shell expansion appears to
   work well. [ BTW: typo "qouted" -> "quoted" ]

2) To follow convention, options should come first:

   Usage: java HTMLConverter [ filespecs ] [-simulate] [-options1 value1
[-option2 value2 [...]]]

   Giving file names as trailing argument appears to work, but
   documentation should follow that.

3) Backup handling is inconvenient:

   a) Documenttation bug:

      source: path to files. Default: <userdir>
      backup: path to write backup files. Default <userdir>/<source>_bak

      This expands to

      backup: path to write backup files. Default <userdir>/<userdir>_bak

      This doesn't make too much sense, regardless of what is meant
      with <userdir>. In fact, backups are created in

`dirname <directory_of_html_file>`/`basename <directory_of_html_file>`_BAK

   b) standard backup files created, even if they already exists and
      contain user (ie, non-HTMLConverter generated) data

   c) with "-backup ." or "-backup `pwd`" (this can happen due to
      buggy Makefiles) no backup is created at all, the source is
      lost. HTMLConverter should reject the case where source==backup.

4) output selection is inconvenient:

   a) even read-only files are overwritten, and are writable
      afterwards; this is annoying, especially in a version-controlled
      environment

   b) source files should always be left untouched (that would free
      from the need for backup files)

   c) output should go to standard output by default

   d) an -o flag sould be provided to explicitly set the output file
      (or output directory if multiple files are given, or output
      directory root if recursion is selected); -o - should still mean
      "write to standard output"

5) HTMLConverter should read from standard input if no files/directories
   are given (have also tried "-source -" which didn't work). Then
   HTMLConverter could be used in a pipe, especially in a Web server's
   conversion chain. BTW, this would be the most useful application of
   HTMLConverter, IMHO.

6) HTMLConverter always looks in its template/ subdirectory for template files. It
doesn't accept relative (to current working directory) or absolute pathnames,
even ./mytemp.tpl, `pwd`/mytemp.tpl or ../../../../../../`pwd`/mytempl.tpl
tricks don't work.

[ PS: there are minor typos in the help text:

template: path to template file. Use default it unsure.
                                               "if"
                                              
log: path to write log. If not provide, no log is writen
                                "provided" "written." ]

7) Although the -help text reads
...
log: path to write log. If not provide, no log is writen
...

a file convertlog.txt is created, even if -log flag is not given.
(Review ID: 101359)
======================================================================