jackrabbit-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Scott Cytacki <scyta...@fastmail.fm>
Subject replicated/synchronized repositories
Date Mon, 16 Jun 2008 20:12:43 GMT
Hi.

I'm considering using Jackrabbit to load and store learner data for 
applications we are running in schools.  I've been experimenting with 
the various ways to remotely access the repository, and so far none of 
them are fast enough and robust enough for my needs.   I've tried
- jcr-rmi,
- jcr2spi -> spi2dav
- jcr2spi -> spi2jcr -> jcr-rmi,
- jcr2spi -> spi-rmi

I want to read, write, and query nodes during a session.  In school 
computer environments the network is typically very slow.  It is also 
quite likely that the network will go down during a class period.   And 
it is likely that the client code will crash.  
So I want something that can work offline if necessary, be saving a 
local copy of the data in case the client crashes, and when the network 
is available sync up a central server.

Has anyone addressed a problem like this with Jackrabbit?

I think the only way to solve it is to have a repository running on the 
local computer and replicate/synchronize it with a remote repository.   
It seems if versioning is used, then this syncing process can be pretty 
efficient because  it will be possible to keep track of which nodes have 
changed.   Then only those nodes need to be synced.  And only the 
changes to those nodes need to be transfered not the whole state of the 
node.   Also, I would hope the existing clone and merge code which is 
meant to work between workspaces can be reused  to deal with conflict 
detection and resolution.

Has anyone attempted this? 

Is there something other than Jackrabbit that would be better suited to 
solve this problem?

Scott Cytacki

Mime
View raw message