IDEAS FOR RYZOM

Being able to forecast the weather with some degree of success would be a nice option to have. Any way we can implement a Meteorology occupation? I'd imagine it would be a high-level undertaking that would carry with it a sufficient chance of failure (since we all know it's not an exact science).

I'd imagine it'd be something one would have to learn from the Zorai (which teaches Magnetic Cartography - perhaps this could be a pre-req?) or Tryker (which is an inventive species that would have a vested interest in at least predicting wind patterns for windmill placement and optimization), but that's just my take. Yours may vary.

Thoughts?

Interesting idea, but I doubt that it is workable since the "weather" as I understand it is based on a random walk of the humidity in a region biased by season and region.

Until the interface shows the correct weather, however, any attempt at forecast is doomed, and the interface does not, in general show the same weather that the resources are showing.

If it *were* possible, I'd see the possibility for at least two options. The Tryker scientific method, the Zorai mystic method, the Fyros method that only predicts fair or cloudy and the Matis method that involves looking at the sides of trees....

For any forecast to be possible server would have to "queue" weather change in advance and that I think is not such trivial upgrade. What could be done is introducing thrird skill branch in harvesting, Atmospheric measurement. That way one would be able to progressively determine current weather more and more accurately.

Then again you can just fire off 500m Tracking and get basically the same result.

The server is actually knowing the next weather in addition to the current one - so it would be possible to predict the next weather for the next cycle. If the weather display bug is really fixed as it should have been.

Jarnys -- the weather bug is not fixed.

The weather changes every few IRL minutes (I haven't figured out how often -- it's probably in the code somewhere). To be useful, weather prognostication would have to extend several Atys hours out, just to make it possible to get to weather dependent nodes. Given that it is subject to variability on the fraction of Atys hour scale, that would seem to mitigate against useful prediction except for general trends, even if the stochastic changes get queued up as Olmo suggests.

So a forecast might go... "Observations of the air conditions in the Lands of Continuity suggest unsettled weather for the next six hours." or "The magnetic observations in Void suggest prolonged periods of Fair weather for the day, followed by mostly Thundery in the night." But there would be periods during the day when it rained, and periods of Fair that night. Not the usual errors that we see in mundane forecasts, but just random jitters.

I'd just settle for the weather recording bug to be fixed so I could tell what the changes actually *were*. As it is, I treat "weather" as a random factor.

Queuing weather changes could be actually pretty straightforward and trivial - circular buffer class keeping let's say 4 weather patterns in store, current and 3 upcoming. Then as weather change occurs, change the head pointer to point at next in queue, let rng create the upcoming fourth weather pattern and be all set. Using circular buffer due speed and storage effectiveness.

If the wetaher needs to be in database as well, the patterns could be lazily written to DB from server memory as the system resource usage allows or as a part of larger DB server batch update.

Non-trivial in the sense that service has dependencies and changing any service like that in server environment isn't something you set up after-lunch hours to and still be able to pick up kids.