C/P/N/Q par vs. seq break-even analysis with 10ms think time

[Aleksey Shipilev]

Hi Edward,

This is not the first time I see this article. Was wondering when I will
have enough public exposure to receive the note from you. ...and here it
is. At the danger of feeding the troll:

On 10/16/2012 09:51 PM, Edward Harned wrote:
> Using this framework for other its intended use will only reveal its
> flaws in ways that embarrass the team.

Revealing the flaws is the road to perfection. If you are saying
anything can be good from the start, then you are on the way to
soul-crushing realization.

And, you are clearly confusing the levels of abstraction. Asking a
ExecutableService to expose the appserver API? Please. IMO, if you want
to have something executing Runnables/Callables/whatnot, TPE and FJP are
the only sane alternatives in the world at this point. If you have
something better (I know you *think* you do), why wouldn't you go ahead
and explain to everyone how much better it is, and gather lots of
opinions, analysis, and general research (which TPE/FJP already has).

Note that since we are all pragmatic here, the *descriptions* are not
enough, you should have the *implementation* which performs up to your
expectations. As the seasoned performance engineer I would tell the
difference between the two is humongous.

So, if you have better alternatives, show 'em up, we will make a few
rounds of rigorous benchmarking to separate the prejudices from the
objective facts. Here's the forest for the lambda prototype:
  http://hg.openjdk.java.net/lambda/lambda

...you are encouraged to go and put in the "non-broken" and "efficient"
implementation in, then we'll talk.

Blowing out hot air "something is broken" wouldn't cut it. Everything
sucks if you look at it long enough, but in many cases there are no
other viable options. Please refrain from waving hands around pragmatic
engineers. That is embarrassing for *you*.

-Aleksey.