[Shape-univ] just a note and a small caveat

[Boud Roukema]

hi Zbyszek, Bartek, all,

On Fri, 9 Feb 2007, Bartosz Lew wrote:

> hi shape-univ,
>
> I see there's a debate going on on the shape-univ about the circles stuff.
>
> I see that Zbyszek has problems with compilations stuff
> /configuring/making/compiling/including dynamical, static libraries/ setting
> pathes/ environ variables etc or whatever
>
> first thing:
> I've been trough this too and I consider this as a serious hole in the
> programming course in the studies at UMK !!! this should be changed as soon
> as possible at the undergrad. level
>
> programming is not only writting one's own code but also (or maybe primarily)
> including smbodys others code/ objects/ libraries and it SHOULD be taught !
> IMHO. and it isn't also writting makefiles, making configuration scripts etc.

As a matter of principle, i totally agree.

In practice, adding more coursework for undergraduates means removing
some other coursework, unless we want them to have even less time free
for thiking and organising their own learning. So that's a political thing
requiring wide discussion and/or something flexible like we tend to do in 
astronomy. E.g. piernix was a single lecture for 2nd year students this year,
in the hope to get them to use gnu/linux much faster.

The second difficulty is how to teach this and be able to make some reasonably
objective measurement of the students' success in learning/understanding the
material.

IMHO it's a highly non-linear and complex process - since part of it is things
like grep  but it's also RTFM, finding good introductory texts on different
programming languages, and also automake+autoconf  which i am still only at
an elementary level of understanding using.

One idea i had (if i get to teach a course on this) would be to use
wims (http://cosmo.torun.pl/wims)  to put random errors in code and or
config files etc. and then the student has to sufficiently well
understand the code and/or config files etc. in order to get them to
compile and/or work.  Since the random error is known to wims, it can objectively
check if the student correctly found the error, and more importantly, each
student has code with an (essentially) unique error, so that students can
freely discuss with each other the errors in their individual versions of the
code, but copying verbatim without understanding fails. In other words, if the
students discuss with each other (or even on forums, mailing lists, etc.) they'll
have to explain and understand, not just memorise answers.

Anyway, that would be a lot of work to prepare...


> second thing - not quite my bussines but:
> Boud (,Zbyszek),
> unless Zbyszek deals with the circles packege only for learning how to handle
> with the software things, you should stop him from doing that !
> It's a WASTE OF TIME since I hear his subject is not topology, but grav.
> lensing and cosmic strings. !
> I see it now how little time there is to do anything on the PHD course, and
> there's NO time for looking around. :(

Again, i more or less agree here, though i prefer the word used in
India - "guide" - rather than supervisor. i cannot "stop" Zbyszek or
any PhD student working with me from working in a particular
direction.  Well, i guess i could make threats of not signing official
documents or writing a bad report, but that's not my style and it's
not going to help anyone.

Rather, i think i can make recommendations, which may or may not be
worth listening to, and be something of a guide.  And i think it's
good that PhD students who work with me can make
recommendations/suggestions to each other as well, especially at this
meta-level. :)

AFAIK, Zbyszek *is* dealing with the circles package mainly as a learning
exercise, not just for him, but also for us as a group. There's only a small
number of people who use the circles package, but general software skills, 
including sending in bug reports and knowing what information you need to
send and where to send it, and especially learning to send it to *public*
bug reporting websites or forums or mailing lists, is something useful.

For the most popular cosmology software (circles is not (yet ;)) so highly
popular, the software forum on http://www.cosmocoffee.info is probably one
of the best places to start.

Also i would probably not use the expression "WASTE OF TIME" - i think the
real question is efficiency in use of time - anything we spend time learning
from gives us some useful learning experience. Deciding in advance what
things are more or less likely to be useful is difficult.

Probably a power law mix of a small number of things learnt in depth and
a large number things studied less deeply during short periods of time is 
what most of us do and/or should do.  The question then is what the slope
of the power law should be and how to be sure that the total time required
has a finite integral.

Since we know that \int_1^\infty x^{-n} dx is infinite when n=1, it's
probably obvious that either the index should be greater than 1, or we
should cut off the integral to   \int_1^\x_{max}  x^{-n} dx . Clearly then,
a small change in  n  or x_{max}  can potentially give more time for relaxing. :)

OK, enough meta-science for now... :)

pozdr
boud