Comment 3 for bug 755642

On Sun, 2011-04-10 at 13:41 +0000, John A Meinel wrote:
> On 4/10/2011 1:28 AM, Martin Pool wrote:
> > ** Tags added: performance
> >
>
> Testament sha1 is currently designed to be an O(tree) validator. So we
> have to load all the inventory and generate it.
>
> The only way to have it ultimately scale is if we could only compute a
> subset of the whole testament (tree based). However, the main problem is
> that it becomes highly subject to how you encode the tree. And we've
> tried to have Testament be decoupled from any given repository
> implementation.
>
> I see 2 options
>
> 1) Tie it to the current tree-based CHK repository implementation.
> 2) Create a new (semi-arbitrary) layout, but make the intermediate nodes
> 'cacheable' and able to be looked up based on some easy-to-compute factor