On Mon, 15 Apr 1996, Harvey Gilpin wrote: > >A possible solution might be to implement posix threads and set a mud > >flag to prohibit movement, the exits command, look <direction>, etc until > >the room rebuild is finished. That would limit the inconvenience it > >causes. > > That's one option yes. Though I still prefer fixing the data > structure to fixing it's consequences. I think no matter how it's done there's going to be some drawback. When you use pointers you're trading either lookup time or some extra memory for the realnum lookup table to get more efficient prototype creation. I think using pointers is an improvement but I'm not sure I'd call it fixed. Just new and improved :) > Besides, I know nothing about threads :) I know just enough to be a little dangerous to myself :) They really are wonderful though. If implemented correctly they would solve many problems with lag in circle and would make the server much more flexible. > >> I implemented this on my mud today, and amazingly it booted on only > >> it's second attempt. Almost everything is working fine. New room > >> creation becomes ridiculously easy, and the whole world is so much > >> more flexible. I highly recommend anyone considering this approach to > >> go for it. > > > >The problem with this is that modifying olc to work with this system > >makes it pretty impractical to release, > > That's not really a problem. Both of us have already released our OLC > code for standard Circle. Like you, I'll prolly only release a bug > fix version and then leave it at that. So now our mud code can go in > a different direction. CircleMUD is, after all, only a starting > point. No not a problem for anyone who's willing to make the conversion themselves, but it would be nice to be able to include a more efficient system with my distribution. I don't like allocating a lot of extra memory that goes largely unused, but it seems to work pretty well. Sam
This archive was generated by hypermail 2b30 : 12/18/00 PST