Estimates the timestamp when the first response byte was received for a network transfer.
When the browser's Resource Timing API provides responseStart, that value is used directly.
Otherwise, a logistic model estimates the latency fraction (the proportion of total
request time spent waiting for the first byte) from the transfer size and total duration.
The model uses a sigmoid function parameterized on a feature combining log(bytes) and
log(totalTime). Larger transfers relative to their duration are assumed to have a smaller
latency fraction (most time is spent downloading), while small or slow transfers are assumed
to be latency-dominated.
responseEnd represents when the fetch promise settled (approximately when HTTP headers
arrived). This provides an upper bound on when the first byte was received — the actual
responseStart is at or before this point.
Estimates the timestamp when the first response byte was received for a network transfer.
When the browser's Resource Timing API provides
responseStart, that value is used directly. Otherwise, a logistic model estimates the latency fraction (the proportion of total request time spent waiting for the first byte) from the transfer size and total duration.The model uses a sigmoid function parameterized on a feature combining
log(bytes)andlog(totalTime). Larger transfers relative to their duration are assumed to have a smaller latency fraction (most time is spent downloading), while small or slow transfers are assumed to be latency-dominated.responseEndrepresents when the fetch promise settled (approximately when HTTP headers arrived). This provides an upper bound on when the first byte was received — the actual responseStart is at or before this point.