First of all thank you for such fast replies Ted and Kingsley!
On 3/14/07, Ted Thibodeau Jr <tthibod...@openlinksw.com> wrote:
It is perhaps worth noting that multiple data spaces can be hosted within a single Virtuoso instance, thus removing the need for the federation on the level you describe here. Each shop-owner creates their dataspace, and the "mall"-owner handles the overall interface, including the SPARQL interface over all hosted dataspaces.
Yes this may be enough for me at this time. I'm only interested on experimenting with different architectures for this kind of semantic web communities. It's essential that each data spaces owner could define authorization rules for accessing their data. From experimentation point of view it may not matter wether the data spaces are physically separated or not. I have understood that each users' tripples could be stored in separate tables in a single Virtuoso instance so that trivial authorization could be handled on the table level by simple SQL grants. (It should also be noted that a commercial enterprise of the sort
described requires a commercial license anyway, whether using the VOS or proprietary distribution...) Also, a single commercial Virtuoso instance could be used to federate over multiple VOS instances (whether licensed commercially or not), through a myriad of ways. This is more complex to implement, but may be an effective and sufficiently low-cost solution.
Do you know of any experiences on federating distributed SPARQL queries with Virtuoso? I would think that federating the default table structure for the RDF quads is somewhat different than federating typical relatioal schemas. Optimal configuration may require more expensive licenses at any or
every level. However, functional configurations may be achieved at low cost, or even at no cost while in development.
I'm not planning any commercial applications at this time. I'll just have to develop further the database implementation of Kaivo. The overal vision is described here: http://iki.fi/juvi//kaivo/ The current database is my own adhoc tripple store which is build on BerkeleyDB. It would be a big improvement if I changed to any SPARQL accessible tripple store because now I do not have any real query language and the tripple store does not even support multiple graphs in one store. Regards, -Jukka
