[
https://issues.apache.org/jira/browse/DERBY-4172?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12702018#action_12702018
]
Rick Hillegas commented on DERBY-4172:
--------------------------------------
How does the following user-visible behavior sound? I think this would be easy
to explain and the backward incompatibilities would be very rare (probably
non-existent) and easy to workaround:
* We could distinguish between logical and physical database names:
o A logical database name consists of the subprotocol plus the
database path. That is, each subprotocol defines a separate namespace of
databases.
o The physical database is an actual database which is opened. The
same physical database may have more than one logical name.
* We would prevent a physical database from being opened under two different
logical names simultaneously. Once a database has been opened under a logical
name, other connections to the database must use that original logical name.
After the database has been closed, it can be re-opened under a different
logical name.
> You can open a read-write connection to a database which was originally
> opened by another thread using the classpath subprotocol
> --------------------------------------------------------------------------------------------------------------------------------
>
> Key: DERBY-4172
> URL: https://issues.apache.org/jira/browse/DERBY-4172
> Project: Derby
> Issue Type: Bug
> Components: JDBC
> Affects Versions: 10.6.0.0
> Reporter: Rick Hillegas
>
> The original connection, using the classpath subprotocol, was supposed to
> open the DATABASE (not the connection) as read-only, according to the
> Developer's Guide section titled "Database connection examples". However, I
> am able to write to this database in another connection which opens the
> database using the default, file-based protocol.
> At a minimum, the documentation is wrong. But the documentation may be trying
> to impose a consistent, easily described model on the behavior of our
> subprotocols. It may be that the behavior of our subprotocols cannot be
> described by a simple set of rules that users can grasp easily.
> To show this problem, I created a database and then moved it into a directory
> on my classpath (in this case, trunk/classes). Here is a script which shows
> this behavior:
> connect 'jdbc:derby:classpath:derby10.6' as conn1;
> -- fails because a database which is opened on the classpath is supposed to
> be marked as read-only
> insert into t( a ) values 1;
> connect 'jdbc:derby:trunk/classes/derby10.6' as conn2;
> -- this succeeds even though the database is supposed to be read-only
> -- according to the Developer's Guide section "Database connection examples"
> insert into t( a ) values 2;
> select * from t;
> delete from t;
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
