Hi Draicone, and everybody
On Mon, Dec 10, 2007 at 10:53:15AM +1000, draicone(a)gmail.com wrote:
On Dec 10, 2007 7:05 AM, Brian Salter-Duke <[1]b_duke(a)bigpond.net.au>
wrote:
1. The ability to display molecular models that can be manipulated
by the
reader. This function is available in an open source application
called
Jmol. A php extension is available to run in mediawiki and Jmol
does run
on, for example, the Jmol Wiki. I have tried to get this added to
Wikipedia, where of course it would also be useful, but the
technical
folks, think the php is not sufficiently free of security concerns.
I do
not know php but I am trying to get this issue solved.
Unfortunately the
author of the Jmol php extension is too busy to do this.
If you can provide us with further details, I can get it started on
the sandbox wiki we have running on the server.
This sounds great. Give me a few days. The Jmol wiki has just been
upgraded and I have not had a chance to check things out there. I will
make sure I am clear about the latest versions of Jmol etc. I will also
point you to where you can see what it does. It involves installing Jmol
itself and then a php extension.
However, the security problem remains. I have discussed this in the past
with Tim Starling - see User:Tim Starling on the English Wikipedia. He
is employed by the Foundation as a Technical support person. He looked
at the php code and said it did not satisfy him on security grounds. He
will not allow it on the Foundation servers until those concerns are
fixed. I do not know enough about it to fix them. I will have
another go at getting someone to address the security concerns.
2. The real innovation of our materials is that
they allow the user
to
prepare data for a variety of different computational chemistry
programs
and then run the data, either directly in real time, getting the
results
back on a web page, or by queueing the job if it is larger and
getting
the results back later on a web page or by e-mail. Our materials
use a
variety of quite complex CGI scripts that handle the materials from
html
forms and menus and then run the code. While queueing the jobs is
necessary in some cases, this is really only to restrict access to
the
disk for storing large temporary files. The CGI scripts prevent the
user
from entering large jobs. In this context "large job" equates to
"large
molecule" with the cpu time rising with some some power of the
molecular
size. For teaching, small molecules only are needed and these take
little
time. When a job is queued, for example, it leaves a URL link on
the
page returned by the CGI script. The user can keep clicking on it
until
it returns more than a header saying try again later.
While currently we use both open source and commercial codes, the
commercial codes can be replaced by open source ones by writing a
few
more CGI scripts.
This use clearly needs a server that can run these CGI scripts and
the
programs they call. It does not have to be the server that handles
the
forms page. For a while we ran one application 3000 km away from
the
main server.
I've got a 40kg server with triple power supplies, dual CPUs, triple
hard drives and dual ear muffs to be able to hear with it all that we
might be able to use :) Let me know if this sounds okay and I'll get a
flavor of Linux up and running on it.
That sounds great. Yes, it would need linux and I would have to send you
a number of codes to install. However, until I know how we would run
them on a wiki, I can not progress this. We could have course have a
link that goes off the wiki to your server for the cgi scripts etc. That
might well be the best way to go.
The simplest method is the Huckel method. The code is my own and I can
release it under a free license. The program is in Fortran and
actually uses some CGI routines in Fortran. I got this running
just for the hell of it. Someone said you can not write CGI scripts
in Fortran as they have to be in C, perl or python. The program is
itself the CGI script. In other cases of larger Fortran programs
from elsewhere, the CGI script is in perl and calls the program after
getting the data together.
The Huckel program is one of three used by our module 3. The second is
also a home written Fortran program. The third is a program called MOPAC
which exists in many flavours but one of them is open source. I would
want to get module 3 up among the first.
Module 2 needs to make extensive use of Jmol. We are currently still
using a product called Chime which no longer works well with some
browsers. I need to convert the whole of that to Jmol and that would be
an early priority. Molecular images in all other modules are already
converted to Jmol. Module 1 is very old and out of date. I will not be
using that module.
Modules 5 and 8 use a program that is much bigger and takes longer to
run. However small molecules are reasonable. This module is about what
is called ab initio quantum chemistry. It gives a choice of two
internationally well know programs. One is commercial. The other is free
but licensed - not open source. There are two open source ab initio
programs and I plan to test one of them on our servers soon. By test I
mean write CGI scripts to run it from web pages. It works on my linux
box at home and should work on the Monash University server.
I have no idea what the equivalent to CGI scripts
is within
mediawiki
and how we can run forms (the quiz extension may be a start here),
but
I'm willing to learn. Whether this is feasable, I do not really
know.
However, I think that what I am suggesting could have wider use.
Wikiversity needs to have pages that are dynamic in ways that
wikipedia
does not need. Web based learning has to be more than reading text
and
looking at pictures on a computer screen. I think this is the
greatest
challenge for the technical people on wikiversity.
One of the main purposes of the sandbox server was to allow us to
break out of the restrictions of MediaWiki. Forms in PHP are dead
easy. Even if you don't have a programming background you could learn
in half an hour. MediaWiki can have forms, but it requires some highly
complex code and a few hacks here and there. We would be developing
real applications to facilitate our learning systems, as opposed to
this reliance on MediaWiki for what clearly warrants a dedicated
application.
That sounds great.
There has been a suggestion that these kinds of
computer
calculations
can be handled by distributed computing using users own PCs, rather
like
the SETI project. These could be usefull for very large jobs in
some
cases on WV, but I do not think they are suitable for my
applications
where we just need a few seconds (maybe up to 100 secs) on a linux
machine. Most of the codes we use are unix based. Distributing
computing
is not required and it would delay getting the response to the
learner.
If it's a few seconds of calculations, why do we need a computer to do
it in the first place? Can't students run the algorithms on their own
machines? In any case, for a few seconds of calculation the overhead
of a distributed task system like SETI is totally out of the question,
a server would be best.
There are three reasons:-
1. These programs are massive, particularly the ab initio codes. For
real research projects they can run for days in parrallel. The time can
be proportion to n^4, n^5, or n^7 (depending on method) where n is a
measure of molecular size. Small molecules run quickly but they still
whiz through a lot of code.
2. We want students to get some familarity with real codes that they
will use later in research.
3. They run under linux. There are a few ab initio codes that run on
Windows but they are complex and I know none that are open source. Some
cost lots of money. The simple Huckel program could be converted to run
on a PC, but it would still be quite a bother for the user to download
it and set it up for a few calculations. The MOPAC code is also very
large, even though it is not very computationally demanding.
I agree about not using a distributed task system like SETI. I mentioned
it only because the last time I raised this, I was told that was the
way to go for this kind of think on wikiversity.
This is all very encouraging. If you are prepared to cooperate on this,
I think I can add some really usefull learning materials to Wikipedia
and perhaps give ideas to others while doing it.
Sorry my reply is rather long.
Regards, Brian.
I would welcome suggestions for dealing with
this.
Brian.
Let me know what you think of these.
Draicone
References
1. mailto:b_duke@bigpond.net.au
_______________________________________________
Wikiversity-l mailing list
Wikiversity-l(a)lists.wikimedia.org
http://lists.wikimedia.org/mailman/listinfo/wikiversity-l
--
Brian Salter-Duke b_duke(a)bigpond.net.au
[[User:Bduke]] mainly on en:Wikipedia.
Also on fr: Wikipedia, Meta-Wiki and Wikiversity