'Lo.

On 2016-08-01T10:00:37 +0100
Timothy Ward <[email protected]> wrote:
> Hi,
> 
> This sounds like an interesting project, and it also sounds like OSGi would 
> be a really good fit for what you want to achieve.
> 
> The descriptions of static and dynamic policies for DS references are pretty 
> good in the annotation Javadoc:
> 
> https://osgi.org/javadoc/r6/cmpn/org/osgi/service/component/annotations/ReferencePolicy.html#STATIC
>  
> <https://osgi.org/javadoc/r6/cmpn/org/osgi/service/component/annotations/ReferencePolicy.html#STATIC>
> 
> Essentially a static policy means that everything gets wired together when 
> it’s available, and destroyed again when it’s not available. This can add 
> some overhead in systems that are rapidly changing (i.e. services are being 
> unregistered and registered frequently) but it does make the code a lot 
> simpler as you never have to worry about services changing underneath you.

Got it!

> You may find that some of your core components do want to be a bit more 
> sophisticated, for example if you want to reference optional services, or all 
> services with a particular interface, then you will either want a dynamic 
> policy, or a greedy policy.

Right.

> 
> One more thing that you may need to worry about in your sandboxing is how you 
> deal with modifications that refuse to release the thread(s) that they’ve 
> asked for when they’re being stopped. At a guess you will need to use at 
> least some of the OSGi API to track which bundle(s) are not letting go.
> 

Yeah, that's one of the two things for which the standard JVM security
mechanisms still don't have a decent answer. The two being heap
exhaustion and runaway threads. At least in those cases, the damage is
limited to the application itself.

M

Attachment: pgpWKY_0yMdtU.pgp
Description: OpenPGP digital signature

_______________________________________________
OSGi Developer Mail List
[email protected]
https://mail.osgi.org/mailman/listinfo/osgi-dev

Reply via email to