ANNOUNCE: xfce4-weather-plugin 0.8.0 released

[Harald Judt]

Am 26.07.2012 09:52, schrieb Liviu Andronic:
> Hello
>
>
> On Mon, Jul 23, 2012 at 11:54 AM, Harald Judt <h.judt at gmx.at> wrote:
 >>
>> - Major rewrite: Migrate to yr.no API instead of weather.com (bug
>> #8105)
>>
> I'm not sure that there is appetite for yet another provider change or
> if it would be desirable at all. It seems to me that one alternative
> to the current yr.no provider could be openweathermap.org , which
> provides all its data under CC-BY-SA 2.0". I have little experience
> with this provider, but I'm bringing it up since one open-source
> Android client [1] is considering switching to it [2].
>
> Regards
> Liviu
>
> [1] https://code.google.com/p/weather-notification-android/
> [2] http://openweathermap.org/example-json
> _______________________________________________
> Xfce mailing list
> Xfce at xfce.org
> https://mail.xfce.org/mailman/listinfo/xfce

Appetite is not very strong currently. The weather plugin still suffers 
from bitrot and needs fine-tuning. What I haven't found out yet is how 
often the forecasts are updated. It seems to be happen more often than I 
thought.

To elaborate a bit about the current provider: It is mainly a forecast 
API, so what you get is forecasts, not the actual weather. You get data 
like temperature for every hour (or three, or six depending on the 
location), and calculated values for these intervals. One problem is you 
may only get data like temperature for three hours from now, so you 
cannot know the current temperature. I'm not sure the original author 
was aware of this when deciding for met.no and starting the rewrite, but 
then the number of possible candidates was not very big either, and I 
don't think it is now. The values for cloudiness etc. may not always 
correspond to the forecast symbol, which shows the weather for the next 
three hours, and cloudiness is given for a certain point in time, which 
may be the start or the end of the interval. To conclude, the plugin 
uses the forecasts as good as possible to provide current values. The 
values are not realtime, but because they are not too far in the future, 
they are good enough and exact most times.

The current description of the plugin is a bit misleading in that 
regard. Maybe it should just state that it shows weather forecasts, that 
would be more correct.

Also, I'm not sure what I prefer: the forecast for the next hour, or the 
actual weather. To have both would be fine of course. But to know the 
current weather, I can look out of the window here, and the details I 
honestly don't care about very much, the temperature usually won't 
change that rapidly in half an hour that the difference is relevant to 
me. What I find to be of more interest is what's the current weather at 
my home village, or at a place where my friend lives etc. In fact, a 
forecast for the next 1-3 hours is really fine, it tells me to not 
forget the umbrella when I leave my home or working place. That's much 
more important to me than knowing what I can see myself in the next 
minute. Other people will have other preferences though.

As for the openweathermap provider, it seems pretty new. Some parts are 
not documented yet, just look at the missing links in "About 
OpenWeatherMap.org" on the bottom of the website.

http://openweathermap.org/maps:
"Actual weather layer for main cartographic services. Online data from 
more than 20 000 weather stations. Current weather data using is 
absolutely free. We receive data from weather stations with no more then 
1-minute delay after taking. These data are displayed on web-site promptly."

Other questions arise:
- How many weather stations are there in the world?
- How good is the coverage?
- In comparison, how many does met.no support, how good is the quality?

IIRC, one point of criticism of the old provider was that data was not 
available for every location. I've looked for some smaller, less known 
cities on openweathermap beta, and no data was available for them. Don't 
know how their JSON API handles this, it might take the nearest 
available then.

Live data and availability of historical data would be nice indeed. 
Maybe we could consider using this as an addition to the met.no API, and 
maybe switching to it in the future. Before doing so, more experience 
with openweather.org is certainly required.

As a side note: Having a single provider is preferable for maintenance 
reasons.

Finally, I think the android application you referred to still uses 
google weather, at least it reported such a link when it was blocked by 
the firewall.

Harald

-- 
`Experience is the best teacher.'