[Cosmo-dev] Cosmo UI perf profiling

[Matthew Eernisse]

Responses inline ...

Heikki Toivonen wrote:
> You might want to consider standardizing on seconds as the measure of
> time (rather than milliseconds). From the users point of view a
> difference of a few milliseconds won't make a difference, and the
> natural variation is so big that you can't really claim millisecond
> accuracy either.

Interesting idea. I just use milliseconds because that's the units 
Firebug was reporting in the profiler. Anyone else have any thoughts 
about whether it's easier to read numbers in milliseconds or in 
fractional seconds?

> In Chandler all performance times are reported at most to one hundredth
> of a second (when reporting official numbers I used to report with three
> significant numbers and scientific rounding but haven't done that in a
> while).
> 

It might be good now that we have some UI testing automation for Cosmo 
to unify the way we do it between the two projects.

On the other hand, testing for the Web UI involves some significant 
difference (e.g., how long it takes to get resources from the server) 
that might mean we should depart from the way Chandler has historically 
done it. Anyone have any thoughts/preferences?

Thanks for the good feedback. :)


Matthew