On Tue, Mar 13, 2012 at 9:32 AM, Angela Schreiber <[email protected]> wrote: > hi marcel > > >>>> therefore i would strongly suggest to separate jcr-transient >>>> space from an "SPI" layer from the very beginning. >>> >>> >>> Yes, I think we all agree on about the separation. >> >> >> actually, I'm not yet convinced ;) > > > just because we never had time to create a native SPI implementation? > the current spi2x implementations all suffer from the having > JCR on the 'server' side again... it was never the intention > to be limited to that... instead it was always clear that the SPI > layer is intended to separate the session level (transient space) > from the workspace layer that also had to take care about > the validation and consistency. > > >> using the JCR API also has major advantages because we can >> reuse existing remoting as is. jcr-server exposes the JCR >> API over webdav, spi2dav exposes it again as SPI if needed >> and you can even stack jcr2spi on top of it again if you >> want JCR remoting over webdav. >> >> doesn't a new Oak SPI mean we have to re-implement >> all these modules again? > > > jcr2spi needs major refactoring anyway. > spi needs adjustments to fit our needs > > and dav remoting on the spi layer could be much better > if we didn't have to convert spi to jcr again. > > >> I'd suggest non-java clients use the following stack: > > >> non-java-client<-> jcr-server<-> jcr2mk<-> mk > > > not really. as far as i know it was > jcr-non-java -> spi -> mk
i agree with angela, i.e. we should definitely separate the jcr-transient space implementation from an 'spi' layer providing support for the jcr workspace operations, node types, versioning, search, batch r/w operations, etc. cheers stefan > > regards > angela > > >> Regards >> Marcel
