view release on metacpan or  search on metacpan

lib/DBD/MVS_FTPSQL.pm  view on Meta::CPAN

Share locks are acquired only for those rows that are part of a result set.
This prevents dirty reads (the reading of uncommitted data) and nonrepeatable 
reads while phantoms phenomena (described below) can occur.
If a query is issued more than once in the same transaction, it may get additional 
(precisely phantom) rows, as another concurrent transaction can insert rows that match 
the search criteria of the query.

=item Cursor Stability (CS)  

This is the default isolation level. It locks only the row (the page) that is currently being returned. As the cursor leaves the row, the lock is released and acquired for the next one, until all the data is returned.
While this maximizes concurrency and prevents dirty reads it does not 
ensure that the data retrieved will not be changed by other transactions, so
if the transaction reads the same row of data more than once odds are it
gets different results each time (nonrepetable read phenomena).

=item Uncommitted Read (UR)

With this isolation levels the transaction (almost) doesn't acquire locks and 
doesn't check if the data that is retrieving is locked.
This, at the price of risking reading non committed data,