Split Capabilities: Making Capabilities Scale

[Karp, Alan]

> -----Original Message-----
> From: Jonathan S. Shapiro [mailto:shap@eros-os.org]
> Sent: Saturday, July 08, 2000 6:36 PM
> To: Karp, Alan; 'Mark S. Miller'
> Cc: e-lang@eros-os.org
> Subject: Re: Split Capabilities: Making Capabilities Scale
> 
> 
> > Lazy creation is a good idea, but if you're talking about 
> objects with a
> > long lifetime, such as files, you'll eventually end up 
> creating all or
> > almost all of the capabilities.
> 
> In my previous comments on this, I failed to make something clear.
> 
> In the course of it's lifespan, every file in a file system 
> will at some
> time have a read-write capability to it. At some point, most 
> of these become
> unreferenced and could in principle be forgotten. This is why 
> I was asking
> about "dropping" capabilities in the previous mail.
> 
> Alternatively, we can imagine these capabilities as being 
> "out of cache".
> The cost of refabricating capabilities that (statistically) 
> will never be
> used goes rapidly to zero....
> 
> This is why I don't think the argument "they will all exist 
> eventually" is
> compelling. It seems to me like a cacheing problem, and I am probably
> failing to see something in the e-Speak design that makes this sort of
> approach undesirable.
> 
> shap
> 

Cacheing is the right way to look at it.  At least the way we implemented
it, getting a capability in e-speak takes a bit of effort, and the Beta 2.2
capabilities are quite small.  Hence, you tend to hold onto your
capabilities longer, perhaps forever.  It's really just a matter of degree.

_________________________
Alan Karp
Decision Technology Department
Hewlett-Packard Laboratories MS 1U-2
1501 Page Mill Road
Palo Alto, CA 94304
(650) 857-3967, fax (650) 857-6278