Re: [Phys-l] uncertainty +- error

[John Denker <jsd@av8n.com>]

On 09/30/2011 03:21 PM, Dan Beeker wrote:

>  Then one fall I had a group of international students in my 
> lab practicum who absolutely would not single line their mistakes but 
> did single line all their uncertainties. It took me a while to figure 
> out what was going on. When I approached them about it they said it was 
> how the lab manual said they should do it i.e.  "single line your 
> errors". They thought it sounded kind of strange but since I was their 
> instructor and I required them to do it they would do so. Never mind it 
> didn't make an ounce of sense (at least to me). All I can figure is 
> their dictionaries translated uncertainty to "error" and "error" to 
> "mistake".

That's funny.  Slightly sad, but funny.
(I guess I have a twisted sense of humor.)


Possibly constructive suggestion:  

I've been keeping my logbooks electronically for many years now.
I'm not saying everybody should do this, but for me it is a huge 
win.

In particular, it makes the "crossing out mistakes" problem go
away, immediately and completely.

Here are some of the details:  I use  git  to do revision control
on everything I care about, including c++ programs, LaTeX files
(used to produce web pages and other writeups), drawings, and
lots of other stuff ... including logbooks.

Git knows the entire revision history.  It can reconstruct any
previous version at the push of a button.  So if I want to 
change something, I first make sure the previous version has 
been committed to the repository.  Then I change whatever I 
like, and commit the new version.  This means the current 
version always has whatever it should have, including all
corrections, but the older versions are immutable and readily 
available.

Git is fully distributed so you can have N people working on 
the same project.  Indeed you can have more than one person
editing the same document, subject to mild restrictions.

You can cryptographically sign a commit, or a version, which is
dramatically better than signing a page in a paper logbook.  My
signature on a piece of paper means I saw the piece of paper;
there is no real way of knowing if something was added or
crossed out after I signed it.  But a digital signature is
tied to the actual content ... and if the slightest detail is
changed, the signature still applies to the original version,
not to the changed version.

I like git because it is general purpose.  There exist other
more specialized electronic logbook apps.  See e.g.
  http://www-ese.fnal.gov/ods/ode/Electronic%20Logbook%20Evaluation.htm

Git was created by wizards for wizards.  It is not overly
easy to use.  I've used it hundreds of times per week for
many years, so I've gotten used to it.

Summary:
 +++ immutable history
 +++ timestamps and signatures
 +++ fully up-to-date current version, with all corrections applied
 -- some usability issues

For my purposes, this is a huge win.