As the property has a default value, so there is a problem to use two properties for the same purpose. We don't really know if an application uses the misspelled name, or intended to use the default value.

For the current applications, if the implementation name get used, okay, they get the expected behavior if we change to use the impl name, and no worries. However, if we change to use the doc name, the behavior get changed, and problems come out.

For the current applications, if the doc named get used. Applications may expect it to work, but actually not. If we change to use the impl name, the application still does not work, no additional risks. If we change to use the doc name, the configuration works but the application behavior also get changed (although it is the expected behavior).

I think the risk is pretty low if change to use the impl name.

Xuelei

On 11/7/2019 8:46 AM, Sean Mullan wrote:
I guess another option is to not change the name that is used in the docs, but change the code to look for both properties, trying the docs name first, and then the misspelled name.

Not great, but probably the safest and least disruptive option.

--Sean

On 11/5/19 8:07 PM, Xuelei Fan wrote:
I understand your points.  Between using the doc name and the code name, I think using the code name is a little bit safer if someone really use the impl name.  However, just a little bit.  I’m open to use the doc name if we could get an agreement.

Xuelei



On Nov 5, 2019, at 4:32 PM, Anthony Scarpino <anthony.scarp...@oracle.com> wrote:

I understand the desire to change this, but are we sure the doc should be changed instead of the property?  I would tend to believe users code to the doc and don’t notice it wasn’t working.   Not reading the source code and code to that implemented name.  Otherwise I’d assume someone would have filed a bug already in the 2yrs.

I don’t want us to support two properties, I’m just not confident which way is right.

Tony

On Nov 5, 2019, at 4:00 PM, Xuelei Fan <xuelei....@oracle.com> wrote:

Hi,

May I have the CSR reviewed?
  https://bugs.openjdk.java.net/browse/JDK-8233652

The system property, "jsse.enableMFLNExtension", was introduced in JDK 9 (See JSSE Reference Guides). However, the implementation code uses "jsse.enableMFLExtension" (without 'N') instead.

As the system property may have been used in practice, it may be better to change the CSR and doc accordingly.

Thanks,
Xuelei


Reply via email to