[sldev] [META][AWG]log chat of AWG meeting Friday, Oct 5, 2007

[Argent Stonecutter]

On 12-Oct-2007, at 02:23, dirk husemann wrote:
> see, i'm assuming that assets don't exist in inventories, but rather
> that an inventory has a reference to an asset "living" in an
> inventory/asset server somewhere.

Replace "inventory" with "the asset servers in the trust domain that  
includes the inventory".

The logic doesn't change.

> if we assume an asset to consist of meta data ("properties") and a
> reference to the data (current asset UUID, if i understand things
> correctly),

Since the permissions are part of the properties, and the UUID refers  
to an object with permissions, the UUID has to refer to the asset,  
not the asset data.

> then giving a copy to someone should make a new copy: the
> meta data part changes, the reference to the asset server might remain
> the same

That's right. This copy is a new instance of the asset. It might or  
might not include the data in the asset, but that's a function of the  
trust model and caching algorithms.

> --- the other avatar might want to take that copy into its own
> asset server, though. if i rezz something that should create a new  
> copy
> as well to decouple it from the inventory object --- otherwise we  
> end up
> with an unmanageable tangled web of references with the region domains
> having to constantly check whether the representation of an asset has
> changed.

The whole point to this model is to avoid having that happen, in an  
environment where trust relationships are an arbitrary directed  
network. The only time you NEED to compare the instance with the  
original is if you have an instance that's got the potential of being  
"saved back", *and* that "save back" operation happens.