On Fri, Jun 06, 2008 at 08:24:47PM +0100, Peter Tribble wrote: > On Wed, Jun 4, 2008 at 11:38 PM, Nicolas Williams > <Nicolas.Williams at sun.com> wrote: > > Below is my first draft of an ARC case for this. Please review. > > > > - svcprop -o sh ... > > > > The -o option selects an output format. > > I'm a little concerned about the choice of -o. To me this means > either that we're specifying an output file, or that the following is > a set of output fields. (And I'm sure it has other menaings in other > cases too.) Using -o in this context to mean something completely > different could be confusing.
-o could mean lots of things (ps -o ...? truss -o ...?). Roland suggests -F. Let me be clear: I don't care which it is :) > > Initially the only output > > format selectable with this option will be 'sh', but in the future we > > may add "-o ksh93-compound", an output format that leverages Ksh93 > > compound variable values. > > Hm. Why concentrate on shells? What about other languages? The idea is to support potentially multiple formats. Initially I care about shells 'cause I'm writing a script around svcprop(1) and svccfg(1M). > config/foo astring "xyz > 123" > > Is there something special about multiline values? The problem here is > that many utilities that you might feed this output to are expecting > to chew lines, so replacing the newline with a literal \n would be a > great improvement. I'm finding lots of problems with quoting and shells. More than I'd thought existed. I have to step back and play with this some more. Nico --
