Hey jonas,
I was very happy about your private mail recently in which you mentioned that you don’t want to write PRD’s but that you really want to start coding. The same holds true for me.
Today I am at ESSIR presenting a poster and graph data base model for social news streams. The advantage of my approach is that it is able to create the newsstreams for any user in a very fast way. That will scale perfectly and is actually independend from the size of our data base! It also uses all the advantages from neo4j in a cool manner.
Latest by friday I will make my slides and the poster I present publicly available and also write a blogpost about this. I would suggest to implement this data base model with our metalcon data base in mind. Thereby creating a newsstream around the following data:
- User
- Band
- Album
- Review
- all relations between those
This is already a very huge project since we have to integrate all the things like discussions, comments, relations between them. Profiles / newsstreams around the band or user and all levles of privacy settings.
Once we are done with that we can easily do some usability testing and see weather our new ideas are being liked and from there start pulling over all the rest.
The advantage of beginning in this way are huge:
- we have already all the technologies (neo4j / gwt) running
- we have a solid data base model that – as you will see – comines the advantages of twitter and facebook approach and has been thought through from several sides.
- it seems to me like a managable project that can easy extend to all the other content (city, location, concert) that we have
I’ve just read your slides and your timing problems reminded me a bit of Lamport timestamps. I don’t know if you’ve already heard about it but probably you can find some interesting ideas in literature about timing in a distributed system. There you have a slightly similar problems when you want to logically order events on several PCs.
Hey I just read http://en.wikipedia.org/wiki/Lamport_timestamps and I don’t see a similarity. I was trying to make this “lamport” approach when building this global interlinked index. But I realized it would break everytime I choose an entity and create the social news stream from its ego network
I didn’t mean that Lamport itself could help you but the whole super computer timing problem could be interesting for you as the problematic seems to be similar.
actually that is true. Since I am supposed to write a paper I need related work and looking at the papers that cited lamport seems to be helpful right now. Therefor your advice turns out to be quite helpful and I am already reading some interesting work.
Pingback: Data structure for Social news streams on Graph data bases