Frederik Ramm wrote: > Hi, > > Karl Newman wrote: >> Why drop them? > > Because it makes documenting what Osmosis does easier, and it makes > understanding what Osmosis does easier for those who start to work > with it now. Keeping everything that made sense at some time in the > past unnecessarily increases complexity. My first reaction was the same as Karl. The named pipe stuff has never been the source of bugs and does give a way for those struggling to get tasks connected properly to "force" them to connect the way they wish. However on further thought I've never used named pipes since the stack connection mechanism was introduced and they're not terribly easy to understand. Internally pipes don't exist anyway, they were always just a logical command line thing to help people visualise the pipeline. So far I've kept the command line very consistent between releases to try to avoid breaking scripts constantly but this might be a good case for simplifying. The command line parsing and pipeline building code has also been very reliable but the code isn't simple.
Anyway, if nobody uses named pipes then I'm not opposed to dropping them. One problem is that I've never had a great feel for who uses osmosis so I don't know the best way of finding out usage patterns. Perhaps I need to introduce some evil phone home functionality providing user statistics ;-) If we do drop them we might have to improve the verbose logging that helps diagnose task connectivity problems. _______________________________________________ dev mailing list dev@openstreetmap.org http://lists.openstreetmap.org/listinfo/dev