Seems reasonable. If it's not possible (yet) to be distinguish between time to 1st byte and time to first chunk then just document what the figure represents.
On 27 February 2016 at 08:08, Philippe Mouawad <[email protected]> wrote: > Hello, > Any feedback ? > > Thanks > > Regards > > On Tuesday, February 2, 2016, Philippe Mouawad <[email protected]> > wrote: > >> Hi, >> Any thoughts on that ? >> Thx >> >> On Wednesday, January 13, 2016, Philippe Mouawad < >> [email protected] >> <javascript:_e(%7B%7D,'cvml','[email protected]');>> wrote: >> >>> Hello, >>> I think we should fix our computing of latency property. >>> >>> AFAIU when I look at HTTPHC4Impl: >>> >>> - sampleStart : there is some processing that occurs after it is >>> called: >>> - There is some work related to caching >>> - HTTPCLIENT PreProcessors are processed >>> - latencyEnd is called too late I think: >>> - It is called after Response Header have been processed >>> - It is called after N bytes have been read, not the 1st byte >>> >>> >>> I would propose the following changes: >>> >>> - sampleStart : Create a RequestInterceptor that computes sampleStart >>> - latencyEnd : Call it in >>> MeasuringConnectionManager#receiveResponseHeader although it is still not >>> the best place >>> >>> I think we need to ask HttpCore/Client project to either compute those >>> and put it in HttpConnectionMetrics or add some hooks so that we can do it. >>> >>> >>> Thoughts ? >>> Thanks >>> Regards >>> Philippe M. >>> >> >> >> -- >> Cordialement. >> Philippe Mouawad. >> >> >> >> > > -- > Cordialement. > Philippe Mouawad.
