Le Jeudi 2 Juin 2005 09:20, Mark Wedel a écrit : > I was thinking about this metaserver some more, and had some various > thoughts. > > 1) Has someone agreed to run the metametaserver? What about the slave > metaservers? While all this looks nice, if you don't have people willing > to run it/take care of it, it is all pretty pointless. For the > metametaserver, I'm thinking more of the Tanner's, since they currently > hold the main crossfire home page, but of course, the metametaserver could > move someplace else (not as sure how many people have as stable and ip > address and network connection that they do). That's were a static list on various website is usefull, easy to place (crossfire.sourceforge.net could be a place to hold the list) > > 2) having the clients connect to the various servers to get stats like > current number of active players is IMO an extra level of complication that > can be done without. If we have a dozen active servers (which was actually > low if I recall), you probably don't want to do it sequentially - that > means it coudl take a significant amount of time to go through them all. > Especially if one is down or unresponsive. this means you need to do it in > parallel, and writing that threaded code is a bit more complicated. Add to > the fact that the client would then have to send some commands and parses > the responses from the server. I just don't see any reason that dynamic > info shouldn't be included. could be done in udp (send a packet to each server than wait 2 second for any response) > > 3) I'd much rather have a simpler approach than something really > complicated for many reasons. The current approach seems to be going > towards the overly complicated approach. probably > > 4) Has anyone considered the approach where all the servers talk to the > metametaserver (lets call it the master metaserver), and all the > metaservers just regularly pull updates from the master metaserver? And we > disallows directy client queries to the master metaserver (basically, it > will only allow queries from the client metaservers)? This goes to #3, > where it seems to be a much simpler approach. And it is only slightly less > resilent then the proposals I've seen. In this approach, if the master > metaserver is down, the clients just use the last copy of the data they > have. But at least all the client metasevers are sure to have the same > data. Could you rephrase it, am not sure to understand (i think there is a mix between master server clients and game clients) > > Just some thoughts. I'd probably still be good to rewrite the metaserver > logic, but IMO, having the servers talk to web servers is an extra > complication - pretty much all metaservers that are out there (certainly > netrek at a minimum) use their own protocol for the servers to update the > metaserver. > I was proposing something that could be run without special servers. A static list anyone could mirror on their isp personal web page. Anyway, if it's possible to have at least 3 or 4 mirrors using a metaserver dedicated protocol, why not. > > _______________________________________________ > crossfire mailing list > crossfire at metalforge.org > http://mailman.metalforge.org/mailman/listinfo/crossfire -- David Delbecq Royal Meteorological Institute of Belgium - Is there life after /sbin/halt -p?