I just discovered the following documentation "gotcha" regarding
sed. (It is in 1.0 but I can't easily check if it's still in -current).

For the description of regular expressions, the sed(1) man page refers
to regex(3) which refers to re_format(7):

sed(1):
Sed Regular Expressions
     The sed regular expressions are basic regular expressions (BRE's, see
     regex(3) for more information).  In addition, sed has the following two
     additions to BRE's:

regex(3):
DESCRIPTION
       These routines implement POSIX 1003.2 regular  expressions
       (``RE''s); see re_format(7).  Regcomp compiles an RE writ-
 
re_format(7):
       A  piece  is  an  atom possibly followed by a single- `*',
       `+', `?', or bound.  An atom followed  by  `*'  matches  a
       sequence  of  0 or more matches of the atom.  An atom fol-
       lowed by `+' matches a sequence of 1 or  more  matches  of
       the atom.  An atom followed by `?' matches a sequence of 0
       or 1 matches of the atom.
[...]
       An atom is [...]
                ,  or  a  single character with no other signifi-
       cance (matching that character). [...]


Here we get to my problem: sed does not seem to support the '?' suffix,
which earlier Unixes I used always did.
I tested this with

	echo "hallo" | sed 's/al?/XX/'

which should of course print "hXXlo" but instead it prints "hallo".
I can get it to behave properly by using \{0,1\} instead (not \?).
Nex/nvi behave the same. grep wants \? or \{0,1\}, egrep works with ?
and no other variation.

A weird thing is that grep also claims to implement basic regular
expressions, but apparently does it differently. I also think that
grep is incorrect in requiring \? in a basic r.e.

Am I going mad or is sed (and grep) indeed incorrect?

-Olaf.
--
___              Copyright 1995 Olaf 'Rhialto' Seibert. All Rights Reserved.
\X/    You are not allowed to read this using any kind of Micro$oft product.