Known bugs in this version:

- Very large text nodes cannot be compressed.
  If the difference between the uncompressed and compressed size
  of a node is more than 64K, compression fails. This is actually
  not a bug in the compiler or the viewer, but rather a limitation
  of the file format. This is usually not a problem unless you write
  real huge text pages (i haven't seen a single file where this limit
  is exceeded). It may affect hypfind however, which writes the hitlist
  to a single page: if you run it to search for text which returns a
  lot of hits, hcp may not be able to compile the generated file.

  Such nodes will now give a warning, and are written uncompressed
  to the output file.
  The 64k limit does not apply to images, which use an extra field from
  the index to store the missing information.

- hcp ignores the -dN option.
  hcp currently accepts this option for compatibility, but does not
  use it. This may change in a future version.

- hcp ignores the -bN option.
  This option is not needed by the current implementation. There
  is no limit on the node size other than available memory.

- when recompiling to ST-Guide format, it is almost impossible to
  determine wether a link was originally specified in the source file or
  generated by the auto-referencer. Symbols are therefore always marked
  as "ar" when recompiling. Note also that symbols are not generated
  at all if the corresponding REF-file is missing. In this case,
  you will have to investigate the generated link-commands.

- Not all viewers currently support lines with arrow endings

- hypinfo.ttp is way too large (~16MB), and may not work on TOS.
  Main reason is the builtin database to identify languages.
