On Wed, May 02, 2007 at 03:25:50PM +0200, u+codalist-p4pg_at_chalmers.se wrote:
> On Wed, May 02, 2007 at 04:32:43PM +0900, Stephen J. Turnbull wrote:
> > That's the deal with hoard, I should think: it's been left in a state
> > where it more or less works, but it's not at all polished, yet.
>  
> Agreed.
>  
> I tried hard to make hoard work for me but it did not scale to the amount
> of files to be hoarded. My conclusion (having talked to Jan)
> was and still is that hoard has to be redesigned to be generally usable.

Hoard does much more than it's name implies. Essentially hoard doesn't
just make sure we refetch files that have been updated on the server, it
is also very closely tied into the cache replacement algorithm.

> The biggest problem with hoard was that it insists on revalidating files
> (keeps rvm in RAM and my link busy)

I think this revalidation happens either way, because the hoard walk
also happens when there is no active hoard profile, it recalculates the
dynamic priority which is used for cache replacement decisions.

Jan