[cap-talk] Streaming capability systems

[John Carlson]

I agree that everything should be packets underneath, for efficiency  
purposes.  In Java
I can open up a TCP socket and send streams of data over it.   Are  
there equivalent mechanisms in Joe-E or E?
Say I wanted to share a capability to read and write to my headset.

John
On Feb 29, 2008, at 5:10 PM, Jed Donnelley wrote:

> On 2/29/2008 3:30 PM, Jonathan S. Shapiro wrote:
>> On Fri, 2008-02-29 at 15:10 -0800, John Carlson wrote:
>>> How does a capability language support streaming--very long messages
>>> where you don't know when the end is coming.
>>
>> I don't think this is any different from any other sort of descriptor
>> system. There is nothing intrinsic in capabilities that demands an
>> RPC-style idiom.
>
> I agree there.  We certainly had arbitrary length "sessions"
> that were negotiated over "control" channels that used capabilities
> (as data to be sure) in our NLTSS system.
>
>> That said, I think that the RPC idioms is one of the big qualitative
>> dividing lines among capability-based designs. Streaming designs  
>> require
>> sessionful interaction, and that makes persistence, setup, and  
>> resource
>> accountability more difficult.
>>
>> None of which is a reason to abandon sessions. It's more a question  
>> of
>> which one to have at the bottom if you want resource accountability  
>> for
>> your system.
>
> I also agree there.
>
> I wonder if the focus of John's question might have been on the
> "language" part of the support.  Are you (John) asking about
> support for streams in E or Joe-E?  I'm not the right one
> to ask in any case, but I ask you to clarify your question.
>
> --Jed  http://www.webstart com/jed/
>
> _______________________________________________
> cap-talk mailing list
> cap-talk at mail.eros-os.org
> http://www.eros-os.org/mailman/listinfo/cap-talk