On 03.08.2017 13:53, Evgeny Kotkov wrote:
> Stefan Sperling <stsp_at_elego.de> writes:
>
>> "local network" is a rather blurry concept.
>> This will often return FALSE for local wireless connections, for
>> example, because of shared medium contention.
>>
>> I agree that 5msec is a well chosen latency threshold.
>> But why not name this function after the question it answers,
>> e.g. svn_ra_serf__is_low_latency_connection()?
> My original intention here was to have a function that can distinguish
> between LAN and WAN, using the information that we have — that's
> latency now and possibly something else in the future. But indeed,
> considering local wireless connections, the name of the function could
> be misleading.
>
> Something like svn_ra_serf__is_low_latency_connection() would probably
> be more appropriate right now. And, we could deal with the additional
> information about the connection to base the detection on, if and when
> we have that.
>
> I'll try to update the function name and the related comments.
It occurs to me that latency on a 10 mbps wired ethernet LAN will be
about the same as on a 1000 mbps wired ethernet LAN, especially if the
packets used to measure latency are small. ... but the bandwidth will be
wildly different.
(Just throwing this into the mix because I can't actually imagine anyone
using a 10 mbps ethernet in a development environment these days, but
who knows ...)
-- Brane
Received on 2017-08-04 13:54:10 CEST