What documentation is fair/legal to use to develop clone software?

[Dave Mason]

I just resubscribed to comp.sources.d to post this question &
discovered the end of a related discussion.  Hopefully I've phrased my
question in a slightly less controversial way.

I am developing a clone of some major AT&T software (exactly *what* is a
secret for now).  I obviously don't want the result legally
encumbered, as it is to be freely distributable (a la FSF or less
restrictive rules).

The question is: What can I base my knowledge on?

Some answers: The X/Open manuals?  The Posix manuals?  The UNIX
manuals?  The Bach book?  Reading the .h files?  The SVID?  Other
people reading the source & answering questions (this seems to be
breaking their trade-secret agreement with AT&T)?  The Lyons book?  Me
reading the sources (I'm sure this is not kosher)?  I've listed these
in descreasing likelyhood of being OK.  To this point I have looked at
the X/Open & UNIX manuals & read the Bach book & very occasionally
looked at .h files.  The X/Open manuals don't give binary encodings...
does Posix? 

For example, suppose I were doing a STDIO?  How would I know the
function names, valid parameters, binary values of #defined values
(such as _IOFBF, _IOLBF, _IONBF in setvbuff(3))?  This last point is
the most tricky - short of looking in stdio.h (or writing a program
that tried random values until it got a result I was looking for) I
can't see how one could get the correct semantics.  And the .h are
clearly copyrighted...

I'd really like to hear from people that have actually developed PD
clones (Henry, Ozan, AST, RMS, other FSF, ...) or other people who
KNOW what they're talking about, rather than from people saying ``I
think the law ought to say blah...''.  That's an interesting
discussion too, but I'm looking for facts.  If FSF has a ``Guidelines
to writing freed software'', I'm sure it would be helpful.

Thanks muchly,	../Dave <mason%tmsoft at toronto.edu>