Server side cache on Yahoo Weather API?

It seems a bug of Yahoo Weather API that every time I send a request I get the same weather. What's interesting design, there isn't any timestamp field in the response to identify the time of the weather.

$ curl "http://weather.yahooapis.com/forecastjson?p=CHXX0099&u=c" -v
* About to connect() to weather.yahooapis.com port 80 (#0)
*   Trying 180.233.115.34... connected
* Connected to weather.yahooapis.com (180.233.115.34) port 80 (#0)
> GET /forecastjson?p=CHXX0099&u=c HTTP/1.1
> User-Agent: curl/7.21.3 (i386-redhat-linux-gnu) libcurl/7.21.3 NSS/3.12.9.0 zlib/1.2.5 libidn/1.19 libssh2/1.2.7
> Host: weather.yahooapis.com
> Accept: */*
> 
< HTTP/1.1 200 OK
< Date: Tue, 31 May 2011 12:58:10 GMT
< Cache-Control: private, max-age=1200
< Expires: Tue, 31 May 2011 13:18:10 GMT
< Connection: close
< Transfer-Encoding: chunked
< Content-Type: application/json;charset=UTF-8
< 
* Closing connection #0
{"units":{"temperature":"C","speed":"km\/h","distance":"km","pressure":"mb"},"location":{"location_id":"CHXX0099","city":"Nanjing","state_abbreviation":"JI","country_abbreviation":"CH","elevation":3,"latitude":32,"longitude":118.80000000000000},"wind":{"speed":6.00000000000000,"direction":"N"},"atmosphere":{"humidity":"22","visibility":12.01000000000000,"pressure":1024.00000000000000,"rising":""},"url":"http:\/\/weather.yahoo.com\/forecast\/CHXX0099.html","logo":"http:\/\/l.yimg.com\/a\/i\/us\/nt\/ma\/ma_nws-we_1.gif","astronomy":{"sunrise":"06:22","sunset":"18:09"},"condition":{"text":"Sunny","code":"32","image":"http:\/\/l.yimg.com\/a\/i\/us\/we\/52\/32.gif","temperature":14.00000000000000},"forecast":[{"day":"Today","condition":"Mostly Clear","high_temperature":13.00000000000000,"low_temperature":3.00000000000000},{"day":"Tomorrow","condition":"Partly Cloudy","high_temperature":16.00000000000000,"low_temperature":5.00000000000000}]}

However, if I request the same interface via VPN, the data returned is just right.

$ curl "http://weather.yahooapis.com/forecastjson?p=CHXX0099&u=c" -v
* About to connect() to weather.yahooapis.com port 80 (#0)
*   Trying 180.233.115.34... connected
* Connected to weather.yahooapis.com (180.233.115.34) port 80 (#0)
> GET /forecastjson?p=CHXX0099&u=c HTTP/1.1
> User-Agent: curl/7.21.3 (i386-redhat-linux-gnu) libcurl/7.21.3 NSS/3.12.9.0 zlib/1.2.5 libidn/1.19 libssh2/1.2.7
> Host: weather.yahooapis.com
> Accept: */*
> 
< HTTP/1.1 200 OK
< Date: Tue, 31 May 2011 12:58:10 GMT
< Cache-Control: private, max-age=1200
< Expires: Tue, 31 May 2011 13:18:10 GMT
< Connection: close
< Transfer-Encoding: chunked
< Content-Type: application/json;charset=UTF-8
< 
* Closing connection #0
{"units":{"temperature":"C","speed":"km\/h","distance":"km","pressure":"mb"},"location":{"location_id":"CHXX0099","city":"Nanjing","state_abbreviation":"JI","country_abbreviation":"CH","elevation":3,"latitude":32,"longitude":118.80000000000000},"wind":{"speed":6.00000000000000,"direction":"N"},"atmosphere":{"humidity":"22","visibility":12.01000000000000,"pressure":1024.00000000000000,"rising":""},"url":"http:\/\/weather.yahoo.com\/forecast\/CHXX0099.html","logo":"http:\/\/l.yimg.com\/a\/i\/us\/nt\/ma\/ma_nws-we_1.gif","astronomy":{"sunrise":"06:22","sunset":"18:09"},"condition":{"text":"Sunny","code":"32","image":"http:\/\/l.yimg.com\/a\/i\/us\/we\/52\/32.gif","temperature":14.00000000000000},"forecast":[{"day":"Today","condition":"Mostly Clear","high_temperature":13.00000000000000,"low_temperature":3.00000000000000},{"day":"Tomorrow","condition":"Partly Cloudy","high_temperature":16.00000000000000,"low_temperature":5.00000000000000}]}

Personally, I think this is something related by Varnish or Squid stuff. Are y开发者_开发百科ou also affected by this issue? Before I can find the official support link, what can I do to expire the server-side cache ?

In your question as posted, the two Curl examples look identical, but I imagine you are noticing that the data returned was different depending on your network path.

There is some level of caching on these JSON responses (30 minutes I believe), but there may be additional caching on your network.

One important issue is the weather forecast data on the backend is updated on intervals, because the forecast itself does not change that frequently.

In fact, the JSON weather interface you're using is not officially supported, but I believe it's just used by Yahoo itself on various pages.

Refer to the Yahoo! Weather page on YDN for the "official" supported interface. By itself this interface is not as useful because it's only in XML format. Instead you could use the YQL console to pass in your location and receive back JSON formatted data.

For example, try this query: select * from weather.bylocation where location="Nanjing, CN". In the results, notice the lastBuildData value which is the last time the forecast was updated:

{
 "query": {
  "count": 1,
  "created": "2011-12-30T20:03:17Z",
  "lang": "en-US",
  "results": {
   "weather": {
    "rss": {
     "version": "2.0",
     "geo": "http://www.w3.org/2003/01/geo/wgs84_pos#",
     "yweather": "http://xml.weather.yahoo.com/ns/rss/1.0",
     "channel": {
      "title": "Yahoo! Weather - Nanjing, CN",
      "link": "http://us.rd.yahoo.com/dailynews/rss/weather/Nanjing__CN/*http://weather.yahoo.com/forecast/CHXX0099_f.html",
      "description": "Yahoo! Weather for Nanjing, CN",
      "language": "en-us",
      "lastBuildDate": "Sat, 31 Dec 2011 3:00 am CST",
      "ttl": "60",
      "location": {
       "city": "Nanjing",
       "country": "China",
       "region": ""
      },
      "units": {
       "distance": "mi",
       "pressure": "in",
       "speed": "mph",
       "temperature": "F"
      },
      "wind": {
       "chill": "25",
       "direction": "90",
       "speed": "4"
      },
      "atmosphere": {
       "humidity": "83",
       "pressure": "30.45",
       "rising": "2",
       "visibility": "0.99"
      },
      "astronomy": {
       "sunrise": "7:03 am",
       "sunset": "5:09 pm"
      },
      "image": {
       "title": "Yahoo! Weather",
       "width": "142",
       "height": "18",
       "link": "http://weather.yahoo.com",
       "url": "http://l.yimg.com/a/i/brand/purplelogo//uh/us/news-wea.gif"
      },
      "item": {
       "title": "Conditions for Nanjing, CN at 3:00 am CST",
       "lat": "32.05",
       "long": "118.78",
       "link": "http://us.rd.yahoo.com/dailynews/rss/weather/Nanjing__CN/*http://weather.yahoo.com/forecast/CHXX0099_f.html",
       "pubDate": "Sat, 31 Dec 2011 3:00 am CST",
       "condition": {
        "code": "20",
        "date": "Sat, 31 Dec 2011 3:00 am CST",
        "temp": "30",
        "text": "Fog"
       },
       "description": "
\nCurrent Conditions:
\nFog, 30 F
\n
Forecast:
\nSat - Sunny. High: 50 Low: 28
\nSun - Sunny. High: 51 Low: 28
\n
\nFull Forecast at Yahoo! Weather

\n(provided by The Weather Channel)
",
       "forecast": [
        {
         "code": "32",
         "date": "31 Dec 2011",
         "day": "Sat",
         "high": "50",
         "low": "28",
         "text": "Sunny"
        },
        {
         "code": "32",
         "date": "1 Jan 2012",
         "day": "Sun",
         "high": "51",
         "low": "28",
         "text": "Sunny"
        }
       ],
       "guid": {
        "isPermaLink": "false",
        "content": "CHXX0099_2012_01_01_7_00_CST"
       }
      }
     }
    }
   }
  }
 }
}

I'm not entirely sure if it's equally applicable here but a common solution to IE's aggressive caching for GET requests is to add a timestamp parameter in the request - since the value changes each time the request is made, each request appears unique.

Choose a parameter name that is unlikely to conflict with the API parameters and set it to the current time each time you make the request and see if that fixes things.