home *** CD-ROM | disk | FTP | other *** search
- From: David J. MacKenzie <djm@eng.umd.edu>
-
- > From: peter@ficc.uucp
-
- > Might it not be time to "push the envelope" as 1003.4 has done, and specify
- > Eric Allman's far superior "parseargs" interface? Getopt really doesn't do
- > that much: a command line parser using getopt is only slightly simpler than
- > one assembled completely out of a nested loop, and it doesn't do anything
- > to help generate usage messages and the like... with the result that usage
- > messages that are out of date are not that uncommon, even in system programs.
-
- > Parseargs also helps by providing a system-independent interface, more so
- > if you use my extended version of the unix driver routine. That way folks
- > who work in other environments will be encouraged to produce programs that
- > follow the P1003.2 interface when compiled under POSIX... and POSIX programs
- > will fit well into VAX/VMS, MS-DOS, etc...
-
- Parseargs has a lot of problems; I looked at it and discarded it. It
- might provide a superior interface to the programmer, but it doesn't
- provide the same interface to the user; that is, it doesn't conform to
- the standard Unix option syntax that most programs use (allowing
- ganging of multiple single-letter options into a single argument, for
- example). Since getopt is an existing-practice de-facto standard, I
- see no justification for trying to push something quite different that
- hardly anyone uses as an IEEE standard. I don't think what 1003.4 is
- doing is necessarily a good idea either. It probably exceeds the
- POSIX charter.
- --
- David J. MacKenzie <djm@eng.umd.edu> <djm@ai.mit.edu>
-
- Volume-Number: Volume 20, Number 39
-
-