Mis-leading Summary statistics

[shanti]

Looking at the code in HAR.py, it appears that the various times being reported in the Summary stats are being computed by simply adding up the respective times across all the resources. This makes it appear that the times are much larger than they actually are.
This is because, when there are 6-8 parallel connections made by the browser, many of these har entries reflect actions that are taking place at the same time. So simply summing up the numbers does not give an accurate picture.
To compute this accurately, the non-overlapping times for each type (e.g. connection time, server time) across the timeline needs to be computed and summed.

[pavel-paulau]

This is probably misleading but apparently not unique. For instance, dynaTrace Ajax Edition used to show the same stats.

https://apmcommunity.compuware.com/community/display/PUB/Best+Practices+on+Web+Site+Performance+Optimization

[mrname]

I would like to utilize HAR storage, but I would agree with @shanti that the total asset load times should be computed in terms of asynchronous loading, as opposed to total load time. I am looking at the code right now to see if I can help out with a pull request. In the meantime (if it is helpful to anyone), I created this python module for analyzing HAR files, which provides total asset load times in terms of browser load:

https://github.com/mrname/haralyzer

[pavel-paulau]

I can help out with a pull request.

Pull requests are always welcome!