From: W. Trevor King <wking@tremily.us>
Date: Sun, 5 May 2013 19:55:58 +0000 (-0400)
Subject: calibcant/discussion.tex: Fill in section on data archival
X-Git-Tag: v1.0~256
X-Git-Url: http://git.tremily.us/?a=commitdiff_plain;h=c5044fd20cea36da02a81609f2ebf40043e266be;p=thesis.git

calibcant/discussion.tex: Fill in section on data archival
---

diff --git a/src/calibcant/conclusions.tex b/src/calibcant/conclusions.tex
index dd2460a..3f29e1b 100644
--- a/src/calibcant/conclusions.tex
+++ b/src/calibcant/conclusions.tex
@@ -10,5 +10,5 @@ by the notice of calibration experts\citep{hutter93-erratum} or
 incorrect formul\ae\ are used during the fitting
 (\cref{sec:calibcant:lorentzian}).  By centralizing calibration
 procedures in an open package, \calibcant\ should both reduce the
-effort needed to calibrate AFM cantilevers and improve the quality of
-the calibration.
+effort needed to calibrate AFM cantilevers, improve the quality of the
+calibration, and ease data sharing and archival.
diff --git a/src/calibcant/discussion.tex b/src/calibcant/discussion.tex
index 721edbd..97e0ef6 100644
--- a/src/calibcant/discussion.tex
+++ b/src/calibcant/discussion.tex
@@ -160,4 +160,31 @@ the calculated $\kappa$.
 \subsection{Archiving experimental data}
 \label{sec:calibcant:discussion:data}
 
-TODO
+Scientific data is not thrown away after analysis.  Organizations may
+have standards for archival, and many journals require supporting data
+to be available on request for a number of years after
+publication\citep{TODO}.  Both the raw data and the experimental
+parameters used to collect need to be preserved, but managing this
+manually is tedious and error prone.  Lab notebooks rarely contain
+\emph{all} of the parameters used to collect and analyze a particular
+calibration.  Data collected with \calibcant\ is saved in
+\citetalias{hdf5} with the full configuration
+(\cref{sec:pyafm:h5config}), bundling all of the information together
+in a single file.
+
+One minor drawback to this approach is that configuration information
+(which is not likely to change often) is duplicated between
+calibration runs.  While this uses some extra disk space, the overhead
+is small.  The full calibration datafile weighs in at $3.4\U{MB}$,
+while the calibration section alone is just $37\U{kB}$ (1\% of the
+total).
+
+Besides the benefit of having a self contained file, HDF5 provides
+efficient support for large arrays of typed data (such as the unsigned
+16-bit values from our DACs and ADCs), which is not possible with many
+other open file formats.  The HDF libraries are supported by the
+non-profit HDF Group with a 20 year development
+history\citep{hdf-group} and many users\citep{hdf-users}.  This
+suggests that HDF will be around for the long haul, and if it is
+eventually phased out, that there will be a number of well funded
+organizations interested in developing migration plans.
diff --git a/src/pyafm/main.bib b/src/pyafm/main.bib
index 8312c77..95971ea 100644
--- a/src/pyafm/main.bib
+++ b/src/pyafm/main.bib
@@ -115,6 +115,20 @@
   note = {Version 1.8.10},
 }
 
+@misc{ hdf-group,
+  author = HDFG,
+  title = {HDF Group History},
+  year = 2013,
+  url = {http://www.hdfgroup.org/about/history.html},
+}
+
+@misc{ hdf-users,
+  author = HDFG,
+  title = {Who uses HDF?},
+  year = 2013,
+  url = {http://www.hdfgroup.org/users.html},
+}
+
 @misc{ yaml,
   title = {{YAML} Ain't Markup Language ({YAML\texttrademark})
     Version 1.2},