connectionchange event properties

connectionchange.js

class ConnectionInfo {
  constructor(media="unknown",
              className="unknown",
              classId=0) {
    this.media = media;
    this.className = className;
    this.classId = classId;
  }
}

var typeGenerator = function(type) {
  var id = 0;
  return function(className) {
    return new ConnectionInfo(type, className, ++id);
  };
};
var e = typeGenerator("ethernet");
var w = typeGenerator("wifi");
var c = typeGenerator("cellular");

// Some connection classes, grouped by type only for readability
// These are all rough, and the cellular group in particular
// omits tons of details. I'm not sure if I should be bothered.
var connectionClasses = {
  ethernet: [
    e("10 Mbit"),
    e("100 Mbit"),
    e("1 Gbit"),
    e("10 Gbit"),
    e("100 Gbit")
  ],
  wifi: [
    w("a"),
    w("b"),
    w("g"),
    w("n"),
    w("ac"), // draft
    w("ad")  // future
  ],
  cellular: [
    c("2G"),
    c("GPRS"),
    c("EDGE"),
    c("3G"),
    c("HSPA"),
    c("3GPP-LTE"),
    c("4G"),
    // FIXME(where does wimax go?)
  ]
};

// Example usage:

// We add a "connectionInfo" property to navigator
if(navigator.connectionInfo.media == "wifi") {
  // ...
}

var fetchHighResCatVideos = true;
window.addEvenListener("connectionchange", function(e) {
  // Use navigator.connectionInfo to determine what to do:
  var ci = navigator.connectionInfo;
  switch (ci.media) {
    case "cellular":
      if (ci.classId < 5) { // worse than 4G
        fetchHighResCatVideos = false;
        break;
      }
  
    default:
      fetchHighResCatVideos = true;
      break;
  }
  
  // Carry on fetching cat videos here.
});

So classId is unique to the media but not unique across medias?

1. I'd cleanup the ethernet examples to just stick with single unit (mbps)
2. Wi-Fi: lgtm
3. Instead of listing out the technology names, just stick to generations.. Each generation has different variants and implementations under 3GPP / 3GPP2. The list below should cover them all.
4. Re: WiMax.. Great question. How about bluetooth, and anyone else? I don't know enough about WiMax to say if it deserves subclasses under it, but I'd just propose to keep it as another top level entry: we're grouping technology by physical layer, and WiMax and Bluetooth are fundamentally different from all others.

  ethernet: [
    e("10"),
    e("100"),
    e("1000"),
    e("10000"),
    e("100000")
  ],
  wifi: [
    w("a"),
    w("b"),
    w("g"),
    w("n"),
    w("ac"), // draft
    w("ad")  // future
  ],
  cellular: [
    c("2"),
    c("2.5"), // GPRS
    c("2.75"), // EDGE
    c("3"), // UMTS, CDMA2000 1X
    c("3.5"), // HSPA, EV-DO
    c("3.75"), // HSPA+, EV-DO Advanced
    c("3.9"), // LTE
    c("4G") // LTE Advanced
  ],
  wimax: [],
  bluetooth: [1, 1.1, 1.2, 2.0, 2.1, 3.0]

@jakearchibald: yeah, that's the idea. Making them globally unique felt like an invitation to the sort of meaingless and misleading cross-class comparisons we talked so much about.

@igrigorik: that sounds good, although I was hoping for descriptive names...perhaps as an additional field? I love the cleanup cellular. Will iterate and incorporate.

@slightlyoff descriptive names would be nice, but it gets a bit tricky.. we would have to standardize those as well. I guess between 3GPP and 3GPP2 there are only so many.. But effectively, you're asking for a subclass of the generation. FWIW, within the generation, there is some variability in technical parameters, but mostly on par, so this is more of a nice to have (I think..).