For one of our websites we have a strange repeat view sometimes.
Sometimes the first request of the repeat view is not the test url.
Example test:
https://www.webpagetest.org/results.php?test=170215_15_25077a622e87048c271929e1b314f9dc
Repeat view for Run 1, 2, 4, 5, 6, 7, 8 are not correct
[attachment=680]
Repeat view for Run 3, 9 are correct.
[attachment=681]
Is the HTML cacheable? That kind of looks like what I’d expect to see happen if it has a short TTL (like 1 minute). In everything besides 3 and 9 it is loading from cache so it won’t show up in the waterfall.
Ah, the HTML has a 1min TTL yes indeed. So that is why I don’t see it in the waterfall, I understand.
Why I was posting this: I am plotting all TTFB for this site in a graph for first and repeat view. I was seeing large burst-outs on some repeat view TTFB metric. See attachment for an example.
So the TTFB for repeat views is the TTFB of the first request that is returned from the server not including responses from the browser cache.
Just a thought: Is this then not a wrong representation of TTFB? Should it not be something around almost nothing (a few ms to get it from the cache)?
[attachment=682]
It’s probably more of an implementation detail but WPT largely focuses on “things that hit the network” and in some of the browsers it won’t even see the things that came from cache and can’t time them.