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

[Argent Stonecutter]

On 09-Oct-2007, at 01:58, dirk husemann wrote:
> master: original data + original meta data on original asset server
> reference: reference to data + copy of original meta data, meta  
> data can
> be freshed (on local domain)
> cache: copy of data + copy of original meta data (in client?)

caches will exist all over the place, even internally in a domain...  
in a region host, in an asset server of a domain the object has been  
carried to, in the client.

> copy: copy or reference to data + modifiable metadata
> master copy: exported modified copy of data + exported modified meta
> data (assuming agent may modify), may update master.

master copy: exported copy or reference to data + modifiable  
metadata, may update master.

> copy: currently when i have a no-modify object i cannot rename for
> example, are you proposing to allow an agent to rename? how about
> permissions?

* Permissions are inherently advisory between domains.
* Some metadata, like owner and group, can always be modified even if  
the object is no-mod.

> - master copy: if it's a copy that's been modified, then that should
> really be a new asset.

No, only the metadata is modified. The master copy is simply a copy  
that has only passed through domains that trust each other to  
restrict metadata changes to reflect the same permissions model.

> how about we just follow what john's been suggesting: asset = data +
> meta data. whoever needs to cache, copies data + meta data.
> whoever wants to modify the meta data does so IFF the permissions  
> allow
> for that. meta data always contains reference to original meta  
> data. if
> i modify the data, i get a new asset.

At no point in any of the quoted material did I suggest modifying the  
data. I was simply describing relationships between references,  
cached copies, and metadata.