jackrabbit-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Michael Wechner <michael.wech...@wyona.com>
Subject Re: How Does Jack Rabbit Generate Its UUID
Date Wed, 09 Feb 2011 22:09:01 GMT
Hi Alex

Thanks very much for this summary.


On 2/9/11 7:46 PM, Alexander Klimetschek wrote:
> On 09.02.11 15:45, "Michael Wechner"<michael.wechner@wyona.com>  wrote:
>> On 2/9/11 2:32 PM, Alexander Klimetschek wrote:
>>> On 09.02.11 15:25, "Carlo Camerino"<carlo.camerino@gmail.com>   wrote:
>>>> is there anyway that I could relate this UUID to the filename?
>>>> how does jackrabbit compute what file name to fetch  based on the UUID
>>>> attribute ?
>>> Why do you want to do that?
>> I could imagine applying JCR to an existing file system (e.g. SVN layer)
>> AFAIK http://jackrabbit.apache.org/jackrabbit-spi.html allows to do
>> something like this.
>> Or do I misunderstand Jackrabbit SPI?
> AFAIU, SPI is mainly a simpler interface than normal JCR, with batch
> orientation, to optimize remoting scenarios. For example, a sample usage
> would be this:
> jcr2spi
> spi2webdav
> <webdav protocol>
> webdav2jcr (webdav servlet with batch support)
> and you could more quickly change out the spi2webdav / webdav2jcr part
> using something else, like spi2myfastprotocol / myfastprotocol2jcr for
> example. Without SPI, you would always need to implement the full JCR API,
> jcr2spi saves you that. There is also a spi2jcr to optimize that on the
> server-side as well, i.e. you'd have a myfastprotocol2spi + spi2jcr on the
> server.
> If you want to wrap a different backend and expose it in a JCR-way, you'd
> need to write the full JCR API. Or you could wrap it under the SPI
> interface and use jcr2spi on top, to let clients use your wrapper with the
> In this case you would not use jackrabbit-core at all (maybe copy some
> code ;-)), but mostly use utility stuff from jcr-commons or spi-commons,
> where large parts of general purpose stuff for implementing the JCR API is
> located. There is no use of things like Persistence Managers, DataStore
> etc., because these are meant for implementing a standalone JCR system
> (even the DB persistence managers only use rdbms to store data in their
> format, not to map existing relational data, a common misconception).
> How you handle for example UUIDs (or better, what you expose as UUIDs) is
> then also a choice of your implementation.
> Regards,
> Alex

View raw message