Hi Bill/all,
with the additional discussion, I would support it given that:
1) We provide a way to bundle those information separate from the application
Some kind of security bundle file; it's creation should be much easier
than simple copying from other projects and this also would remove the
requirement to rebuild the complete application. And I am open to the
kind of packaging happening here. It might be in an EAR but also some
additional/separate jar could work in this case.
Maybe it would be a good idea to have some signing mechanism in place
where a "PaaS Account Manager" needs to sign those security bundles
and binds them to specific application names? Could also be part of
the deployment?
2) have password strength requirements for the keystore password (to
make this solution as secure as possible)
Thanks,
- M
> On 5 April 2012 02:20, Bill Shannon <bill.shannon_at_oracle.com> wrote:
>> Have people had a chance to reconsider this proposal in the light of
>> the discussion that followed? The initial reaction seemed to be
>> "it's evil, don't do it". I'm hoping that the discussion clarified
>> our goals with this proposal and you understand better the use cases
>> it does and doesn't address.
>>
>> We'd love to hear the current thinking of the expert group on this topic.
>>
>> Thanks.
>>
>>
>> Bill Shannon wrote on 03/09/12 11:02:
>>
>>> Jason T. Greene wrote on 03/08/12 22:42:
>>>>
>>>> On 3/8/12 6:09 PM, Bill Shannon wrote:
>>>>>
>>>>> I've uploaded another proposal from our security team. Please review
>>>>> and give us your feedback.
>>>>>
>>>>>
>>>>> http://java.net/projects/javaee-spec/downloads/download/credential-ssl-config-ee7-proposal.pdf
>>>>>
>>>>>
>>>>>
>>>>
>>>> Frankly the whole idea of sticking private keys and password databases in
>>>> deployments seems like a major hazard. Developers are used to copying
>>>> these
>>>> around everywhere. I could easily see someone forgetting they have
>>>> sensitive
>>>> information in here. People also tend to use short and bad passwords in
>>>> keystores which makes bruteforcing a PKCS12 file not that difficult.
>>>
>>>
>>> Note that we *already* allow you to include clear text passwords in your
>>> code.
>>> That's nothing new. As always, you have to apply judgment when using these
>>> mechanisms.
>>>
>>> The point of this proposal is to allow you to include passwords in a more
>>> secure manner. The passwords would be encrypted, not clear, and the
>>> encryption password would be supplied separately. If you think we should
>>> have password strength requirements for the keystore password, we could
>>> talk about that.
>>>
>>> Developers already complain that portability of their application suffers
>>> greatly once they have to deal with security. Every product provides a
>>> different mechanism for configuring securing, setting passwords, etc.
>>> It's impossible to write a tutorial that tells people how to write
>>> portable and secure *Java EE* applications because you quickly have to
>>> delve into the intricacies of configuring security for GlassFish or
>>> WebLogic or JBoss or WebSphere or ... Because of that, too many people
>>> just ignore security, or suffer from poor and inconsistent advice for how
>>> to secure their applications. We should be making it easier for people
>>> to write secure applications for the Java EE *platform*.
>>>
>>> Also, remember that we're targeting PaaS with Java EE 7. In a PaaS
>>> environment it's much less likely that you'll have direct control over
>>> the admin capabilities of the underlying app server. And there's
>>> probably not a system administrator that you can appeal to to set this
>>> up for you.
>>>
>>> Remember also that we're trying to create a platform that is both easy
>>> for developers to use when getting started *and* scales to enterprise
>>> deployments. If we *only* allowed things in the platform that made
>>> sense for large enterprise deployments, we would lose lots of developers.
>>> That's why we allow passwords to be included in applications at all.
>>> It's not because we expect large enterprise deployments to have
>>> applications
>>> with clear text passwords embedded all over them.
>>>
>>> This new proposal tries to bridge the gap between developers and
>>> enterprise
>>> deployers by providing the convenience of bundling passwords with
>>> applications
>>> but doing so in a way that provides some security for those passwords.
>>> Rather than telling developers "don't include passwords in your
>>> application,
>>> you figure out what to do instead", we can tell them "here's a secure and
>>> convenient way to include passwords in your application".
>>
>>