applicability of TEMP_TOL to thermostat regions...

[Axel]

On Mar 13, 4:04 pm, Teodoro Laino <teodor... at gmail.com> wrote:
> On 13 Mar 2008, at 20:28, Axel wrote:
>
> > it is hard to let go of something that has been
> > around for a very long time. after having slept over
> > the choices, i'd rather go with a change in the documentation,
> > i.e. keep TEMP_TOL 'as is' and add two remarks stating that this
> > is global only and that regional recaling should use CSVR with
> > short tau instead.
>
> Where does evolution and innovation fits in your remark?

by encouraging people to try out something new without
forcing them to do it right away you have some evolution,
albeit more adiabatic.

> I honestly don't agree with your point.. habits can be (and must be)
> changed and adapted to something different (possibly better).

it may be that with increasing age one gets more prone
to be conservative and also occasionally tired of fighting
for something that is a rather small issue (from the perspective
of a user).

> Due to the continuous computational challenges and growth in
> algorithmic developments, all codes have necessarily
> to evolve into a direction where the organization of the code itself
> has to be revisited, restructured and improved (even when a structure
> was planned since the very beginning).
> The price for not doing that is just an incredible amount of time
> lost in code-maintenance and an incredible increase of the
> difficulties in expanding the code capabilities.

i don't mind people breaking with backwards compatibility,
if it is for a _good_ reason. the problem is that people have
different definitions of "good reason". with a code the size
of cp2k i consider consistency and maintainability _very_ good
reasons. i would rate them even higher than performance or number
of (current) features, as the future potential depends on it.

as you know from the (non-cp2k) stuff i was working on a few weeks
ago, you can easily reach a point where adding a new feature can
become a nightmare, if consistency and maintainability have been
lost during the course of development.

> You touched the real problem. People fight because you move/delete/
> reorganize keywords.. they just see that this creates them the waste
> of (maybe) 5 minutes in order to update their mental scheme. They
> don't see the necessity of these reorganization, They don't see the
> time you spent when you've to modify the code. They want just to
> find the same keywords (used maybe the last time in the 80s) at the same place.

in my opinion, the trick is to find the right balance between
changing something and keeping things the way they were. and
people on both "sides" have to acknowledge the usefulness of
the other view, despite the occasional pain and frustration.

> I wonder: for someone that consider him/herself a scientist shouldn't
> a flexible mind be the key to do research?

absolutely, but then again, to be flexible requires that you
practice a lot and become and expert in the tools you are using.
particularly with scientific software development this is a tricky
issue. you usually don't get paid for the software development, but
for the science you are doing with the result and since few scientists
are really trained in software development/maintenance one is prone to
make all kinds of mistakes in this and quite a few finally adopt a
view
that doing a minimalistic job is the best solution, as it minimizes
the
"pain" of having to work with a tool they are not that comfortable
with
in the first place. this is also the way how you get the most "out of
it".

of course, this logic neglects the effort from others that have
invested
into the infrastructure and particularly in the case of cp2k this
is the part, i appreciate the most because this is where the
potential of the code is: not the individual features, but the
flexibility of how you can combine them and adapt to new problems,
that nobody thought about when implementing the features.
this encourages creativity and to me this is about as important
as flexibility in a scienfic mind.

> The controversial is purely philosophical (don't take me too
> serious). Of course you can change the documentation (I won't do
> that ;-) ).

it is a point that is worth spending a few thoughts
on every once in a while. :)

cheers,
    axel.

>
> Teo