On 9.3.12 14:55, Jukka Zitting wrote:
Hi,
On Fri, Mar 9, 2012 at 2:43 PM, Michael Dürig<[email protected]> wrote:
Why do we need an SPI?
I think the main difference is in whether we want to have all content
accesses go through JCR or not (compare the overview diagram I sent
earlier to the alternative one at [1]).
Having everything go through JCR adds a bit of overhead (stuff like
namespace mappings are normally not needed) and makes it hard or
impossible to express certain kinds of features (like for example
ETags, batch operations or alternative query models).
I'd say that the main design difference between such an Oak API and
the existing Jackrabbit SPI should be that there's no need to be able
to layer the Oak API on top of an arbitrary JCR implementation. Thus
we can better expose and leverage the specific features of an Oak
repository.
Yes this makes total sense. My point was, if the oak-spi is nothing more
than jcr modulo transient space, why would we need another API?
Michael
[1] http://people.apache.org/~jukka/2012/oak-overview2.png
BR,
Jukka Zitting
