At 08:09 PM 6/2/97 UT, Gary Acord wrote:

>we should also talk about backing up data files if we launch into digital
>disasters.  anyone that is depending on there database should be keeping
>regular backups of ALL of there data files, whether it's just one Access file
>or  a hundred MS SQL Server tables etc.
>my general recommendation is one tape(or better media) for each day of the
>week.  and be very religious about changing it each day.

Gary is right, of course, when he says that a rigorous ritual of backing up
data is the best protection against digital disaster. Consequently, he
implies that the need to depend upon the accession sequence identifiers of
objects to create joins may be less important than I assume.  But in the
real world of catch-as-catch-can computerization, in hundreds of small
museums, where computerized collection management is not implemented as
enterprise applications, where mainly one person has total responsibility
for the management software in addition to his/her own job, backups are not
executed daily or (truth to say) even weekly.  In these cases having a
coded accession number to use as the linking field can be quite useful.

Furthermore, backups are not fool proof.  It is not uncommon to have a
backup that will not restore.  Other technical problems can intervene too.
Last month, when my 9gig hard disk decided to fall apart, the first thing
it took with it was Windows 95.  Unfortunately my backup software only runs
in Windows, so the operating system had to be rebuilt by hand before any
data could be restored.

The advantage of using link numbers with no content -- a linear series, for
example -- is that they never have to be revised.  Such numbers carry no
information and only serve to identify records uniquely.  The standard
accession number used in most museums, while encoding the sequence of
accession and separating the elements in an accession, actually conforms to
this requirement except that it documents the history of a museums
accessions and provides a key that many curators and collection managers
can remember.  Our social security numbers are a much better way of
identifying people uniquely, but we still use names in both familiar and
formal situations despite the possibility of names not being unique.  When
important numbers have no meaning at all, it is easier to insert errors.
Meaning provides an added level of error checking.  It is only when
accession numbers attempt to encode curatorial decisions or attributions
(department, country, media, date, etc.), i.e. elements that are subject to
change, that they should not be used for links in relational databases.

R.Baron