Defects in ciforth per 5.1
No known defects.

        Defects in ciforth per 5.0.0 beta

- The save-system for linux 64 bit generates not-fully elf format files,
  though they run.
- The double comparison in forth.lab are wrong: use these
   : D0<   NIP 0< ;
   : D<   ROT 2DUP <> IF > NIP NIP ELSE 2DROP < THEN ;
   : D>   2SWAP D< ;
   : D>=  D< NOT ;
   : D<=  D> NOT ;

        Defects in ciforth per 4.0.7

See defects in 4.0.6. The missing @'s in the documentation is
fixed since 4.0.6. Also the memory i s enlarged, those are the 
only changes.

Severe:  (Always been there, but now the problem is clear.)

The redirection on XP doesn't work for the command interpreter:
it takes over the file from which commands are read unconditionally,
and doesn't head the /c option and the commands after it.
This was not easily detected, because most of the file was already
buffered, and only the remainder of the file is passed to the command
interpreter. If the whole file is buffered, there is no problem.
The effect is that we are left in an interactive shell.

Less severe.
*    Some @ are missing in the documentation.
*    REQUIRE @+ fails
     Work around: precede by `` REQUIRE ALIAS ''.
*    The file rw-line.frt needed for toblk.frt fromblk.frt is missing
     from the generic distribution.
*    A problem imposed on us by makeinfo with the documentation
     system. An apallingly large number of Forth names, can't
     be node names in makeinfo. This leads to the worddocsafe
     macro besides the worddoc macro. We need either
     a Forth makeinfo clone (the preferred solution), or I must
     in the first m4 inversion sneek in the worddocsafe macro
     based on a criterion what characters are present. Oops. I
     just discovered that : (colon) cannot even be looked up
     via de index.

*    make
     as is needed to build documentation generate an alarming number
     warnings especially

*   _ISO_IN_ version cannot load blocks.

*    _ISO_IN_ must be cleaned up to the point that >IN
       has the ISO behaviour, i.e. it can be written.

*     Add an option to put the regression test as examples to the manual.

*     It may be that in some 16 bit version input buffer falls one byte
          short of iso.

*     Real mode booting version no longer works.