[funsec] simple question

Tomas L. Byrnes tomb at byrneit.net
Sun Dec 6 22:46:24 CST 2009


I used unconverted assignments on Digital Research f77 under CCP/M 3.1d
on iAPx 286 chipsets with regularity, and effect, in the early '80s.

 

And after that, I was thankful to never use Fortran again.

 

The bigger point is that the code is garbage, the data not much better
(at least according to the comments, because we can't see the data), and
the researchers have clearly been actively hiding the facts from public
view.

 

It's high time for the Open Source and Free Software ethos to dominate
something on which so much of the future of mankind rests.

 

I, for one, donated lots of CPU time to the BBC climate modeling BOINC
project. I think the idea that there isn't enough computing, never mind
brain, power out there to do this right is complete bunk.

 

Let the science produce the result it will, whatever that may be, but
let it at least be proper science, with the best current practices in
all relevant fields being applied. Then, after the climate models are as
near to unimpeachable as can be (and models can do pretty well, as the
auto makers have shown), we can have the debate about the costs of
various courses of action relative to their benefits and risks.

 

Until we have a model that would pass muster for simulating the
Coefficient of Drag of an automobile (and as far as I can see the CRU
climate model doesn't), how can we base any major public policy
decisions on it?

 

 

 

From: Martin Tomasek [mailto:tomasek at ufe.cz] 
Sent: Sunday, December 06, 2009 4:55 AM
To: Tomas L. Byrnes
Cc: FunSec [List]
Subject: Re: [funsec] simple question

 

Tomas L. Byrnes napsal(a): 

Yes, and it you look at the real to integer, and integer to real,
assignments without conversion functions, you can see where the actual
assignment may be indeterminate. I don't have the datasets, and haven't
done Fortran since F77, but when I did Fortran, assignment without type
conversion produced unpredictable results. Specifically, in little
endian machines, assigning a real to an integer was a very workable
random number generator.


I learned fortran a few years ago, because I needed to test something on
an astronomical library. I used g77 and never met this issue. 

HARRY's docs mention following compilers:
compaq f90, portland group pgf90, g77, gfortran.

And BTW it's depressing to read READMEs. That guy suffered when he tried
to work with old fortran code and datasets.. He had to reverse-engineer
undocumented binary formats, handle nearly-duplicate entries, found out
that someone fixed documented error in the data - but kept same
timestamp on the file. He had to handle weird scaling of coordinates,
obscure measurement units etc.. Search for 'cries' and for '% x10'.




 

 

From: Martin Tomasek [mailto:tomasek at ufe.cz] 
Sent: Saturday, December 05, 2009 4:04 PM
To: Tomas L. Byrnes
Cc: Drsolly; funsec; RandallM
Subject: Re: [funsec] simple question

 

Tomas L. Byrnes napsal(a): 

Hence we return you to my focus: the code. Code does not lie, it merely
does what it was told. 

 

By my reading, the CRU code produces ever more excited random numbers.
Someone else please read it and prove me wrong.


Quote from docs: "Had to make some changes to allow for the
move back to alphas (different field length from the 'wc -l' command)."

If their code depends on calling external programs, it will be difficult
to verify. Unless you are sure of actual output formats. :-)

-- 
Martin Tomasek





 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://linuxbox.org/pipermail/funsec/attachments/20091206/f603c3b0/attachment.htm 


More information about the funsec mailing list