review request for 8010416: Provide a way for DriverManager.deregisterDriver to notify the JDBC driver that it has been deregistered.

[Alan Bateman]

On 19/04/2013 18:34, Lance Andersen - Oracle wrote:
> Hi,
>
> We have been asked  by a few JDBC driver vendors  to allow a JDBC driver to be notified when/if it was deregistered via DriverManager.deregisterDriver if desired.
>
>
> The webrev can be found at http://cr.openjdk.java.net/~lancea/8010416/webrev.01
>
This looks much better than the original proposal, it would have been 
just too problematic to have Driver define a deregister method. Also the 
proposed wording to specify Driver-implementation specific behavior when 
the Driver or Connections is in use, or subsequent use, addresses the 
points that I brought up in the original thread here (thanks!).

Driver

- {@linkplain DriverManager.deregister} -> I assume this should 
DriverManagert#deregisterDriver

- minor alignment issue with the <p> tag.


DriverManager

- one point that isn't covered in the spec is whether the DriverAction's 
deregister is invoked before or after it is deregistered. This 
distinction is probably only interesting for the case that the 
deregister method fails with an error/exception but it's not clear if 
the driver is still registered in that case. For completeness then the 
spec should probably say that any error/runtime exception is propagated 
to the caller of deregisterDriver.

- could the new (and the original) registerDriver methods specify the 
behavior for the case that the Driver is already registered? This brings 
up the question as to whether the DriverAction is overridden if already 
registered.

- the @param alignment is inconsistent in the new deregisterDriver.

- the re-wording of the original deregisterDriver looks much better. 
Minor nit: "was null" -> "is null".

- DriverInfo - would be cleaner to extend the constructor to take the 
DriverAction. Also an action() accessor would make the usage a bit 
cleaner too.


That's all I have for now.

-Alan