jackrabbit-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alexander Klimetschek <aklim...@day.com>
Subject Re: [jr3] Plugin architecture
Date Thu, 18 Feb 2010 13:00:06 GMT
On Thu, Feb 18, 2010 at 13:53, Guo Du <mrduguo@gmail.com> wrote:
> On Thu, Feb 18, 2010 at 12:07 PM, Alexander Klimetschek
> <aklimets@day.com> wrote:
>> The import/export XML methods are part of the JCR spec, so we cannot
>> easily change their well-defined behavior. Instead, to support large
>> dumps or backups, we should define an API at a lower level, that also
>> ignores type and constraint checking to improve speed (if you know
>> your dump is valid), like it is possible in many RDBMS.
> I mean JCR aware import/export, data could be moved from one
> repository to another as a whole or partial, their back end persistent
> manager could be different.

That import/export I am talking about would be part of the microkernel
API, so it would work on all upcoming persistence layers in

Having it completely on the JCR level makes it slow by definition,
since then it would go through all consistence checks defined by the
spec. And you could not import versions, as they are immutable from
the point of the JCR API.

The RepositoryCopier we have now works on the persistence manager
level and does exactly that. But in jr3, this API level will change.


Alexander Klimetschek

View raw message