First of all, the JS scheduler has a certain granularity - that is, you can request an interval smaller than, say, 20 msec, but it will not fire immediately - what you could see is 20 events fired off every 20 msec.
Second, even if you could, this is not a good idea: you would be making 1000 requests every second, from every computer which uses this script. Even if the client and their connections could handle this, it's nothing short of a DDoS for the JSON server.
What you could do is this:
- get time from JSON-NTP (once), this will be a Date
- get local time (once), this will be a Date
- calculate the difference between NTP and local time (once), this will likely be the number of msec that local time is off
- for every time calculation, take the difference into account
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…