On Wednesday, 14 October 2015 at 15:54:49 UTC, Laeeth Isharc
wrote:
For a long time, Ola, I am done discussing with you. But I
would ask you to take more responsibility for the effect of you
words. The piece you quote is from the question, and not from
what I wrote. You refer to it as flame bait, and don't make
this clear in what you write.
The question as phrased is flame bait and trolling because the
answer is obvious and generates lots of noise for no reason.
Latency is not affected by a script that does very little work
compared to all the other causes for latency in a complex
architecture. Including memcache access, database retrieval and
compression.
What is affected by using Python over Go/D is the number of
instances that run the service.
But for a well designed architecture up to 99% of the work is
done by specialized infrastructure, implemented in
Java/Erlang/C++, that often is too costly to develop for a single
project. So you use ready-mades.
Low latency and scaling is a result of architecture, not brute
force computation.
