Link Walking and Security

José Ghislain Quenum jose.quenum at
Wed Jan 21 07:23:57 EST 2015

Thanks Chris for your email. MR is the way I do my link walking actually.
However, I was assuming that the links entry in the meta part of an object
was going to disappear. Please correct me on that if I am wrong. Otherwise
I should be just fine then.
On the other hand I did consider using the new data types, but then found
out that they were not compatible with functionalities such as 2i, etc. Any
pointers on that?

On Wed, Jan 21, 2015 at 2:03 PM, Christopher Meiklejohn <
cmeiklejohn at> wrote:

> On Jan 21, 2015, at 12:21 PM, José Ghislain Quenum <
> jose.quenum at> wrote:
> well the application I am developing is Collaboration support tool. You
> have a bunch of concepts (organization, project, collaboration activities,
> events, etc). Some of these concepts are connected. So, I could have
> modeled them with a Relational DB. But I really wanted to take advantage of
> Riak's distributed nature. So I decided to use links to model the
> connection between the concepts. Now transitioning to Riak 2.0 is a bit of
> a problem to me. Since I need to change the design of my DB and remove
> links and their walking. I have been considering moving the links to a
> graph DB like Neo4j or even redesign the links as additional attributes in
> the concepts themselves. None of these seems like a clean solution to me.
> If I could have pointers about new ways of designing connections or links
> in risk 2.0 without using link walking that'd be cool.
> Any help?
> José
> Hi Jose,
> As Brian Roach notes here, [1] you can still perform link-walking through
> the MapReduce, which is how the functionality has always operated when
> using the Riak Protocol Buffers interface, given it never supported this.
> This should make it slightly easier to upgrade your application to 2.0;
> although you won’t be able to use the deprecated client interfaces.
> That said, you may want to investigate using the new Riak data types [2];
> if you’ve already adapted your application to deal with eventual
> consistency without any specific causal guarantees (independent object
> updates — possible to update a link and have delayed propagation of the
> linked object; hence, no causal consistency), you could investigate
> alternative ways of representing a graph such as using the map data
> structure to track edges and vertices, etc..  Obviously, this is
> application specific based on your consistency requirements and barring any
> size limitations of your data set.
> Thanks,
> - Chris
> [1]
> [2]
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>

More information about the riak-users mailing list