Wrong bodySize (and feature request: add _transferSize)

I’m test-running timing-fix branch code, found there are still a few fields missing from this chrome-har-capturer than a real Chrome DevTools, one is request headers, I can only see Referer and User-Agent while a real Chrome DevTools gives all headers sent in full, like Host: ..., Accept-Encoding: ... and others, I haven’t looked at the chrome remote debugging protocol, not even sure if remote-interface ever provided those, or can it be done in this chrome-har-capturer ??

          "headers": [
            {
              "name": "Referer",
              "value": "http://www.podcastone.com/widgetrecent?progID=877"
            },
            {
              "name": "User-Agent",
              "value": "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/52.0.2716.0 Safari/537.36"
            }
          ],

Another one I found from a real Chrome DevTools is the _transferSize; when in transfer-encoding: chunked mode, the response is (a length + a chunk) + (length + chunk) + … + 0; so the real transferred bytes on the wire could be slightly larger than bodySize; I’m not sure which har viewer ever take use of the _transferSize but it can be used in real bandwith calculation or something, and it’s missing from this chrome-har-capturer

          "content": {
            "size": 118465,
            "mimeType": "text/html",
            "compression": 96200
          },
          "redirectURL": "",
          "headersSize": 662,
          "bodySize": 22265,
          "_transferSize": 22927

[cyrus-and]

So to summarize, the header problem is fixed now. While the bodySize (i.e., the on the wire body size) is reported incorrectly by Chrome. For example:

$ curl -s -H 'Accept-Encoding: gzip' http://example.com/ | wc -c
606

Which is also the value reported in the Content-Length header, is the correct one. While the HAR reports:

"headersSize": 372,
"bodySize": 978,
"content": {
    "size": 1270,
    "mimeType": "text/html",
    "compression": 292
}

Compared to the HAR generated by DevTools directly:

"headersSize": 372,
"bodySize": 606,
"content": {
  "size": 1270,
  "mimeType": "text/html"
  "compression": 664
},
"_transferSize": 978

Note that in both cases compression satisfies:

compression = size - bodySize
292 = 1270 - 978
664 = 1270 - 606

[cyrus-and]

OK, fixed. It seems that encodedDataLength represents the encoded payload plus the headers. I also added _transferSize.