Alexander Klimetschek wrote:
On Thu, Feb 25, 2010 at 13:28, Michael Dürig <michael.due...@day.com> wrote:
I like this approach. In general I think merging is very difficult if not
impossible at all to get right and will always be based on some assumptions
(intended semantics). I think we should not put this burden onto the
micro-kernel but rather leave it to higher levels (or even end users) to do
the/some merging.
Doing so wouldn't prevent a specific micro-kernel implementation from
still trying to do a merge? Because I think this could still be
interesting when following eventual consistency models with
distributed cluster nodes.
Sure but I think we should restrict this to the cases where 1) the merge
semantics is 'obivious' and 2) the performance hit is neglectable.
Michael
