[RFC] Invite Razor-Qt to next Workspaces Sprint

[Vishesh Handa]

On Sun, Feb 17, 2013 at 5:04 PM, Ivan Čukić <ivan.cukic at kde.org> wrote:

> On Sunday, 17. February 2013. 12.20.00 Aaron J. Seigo wrote:
> > On Sunday, February 17, 2013 08:55:38 Martin Graesslin wrote:
> > > * which pulls in Virtuoso
>
> Just a small off-topic here: Virtuoso *is* one of the fastest rdf databases
> out there. The problem is not that much with its speed, but more with the
> graph database searches for any database* - just imagine a normal
> relational
> database with a join for every sentence in sparql.
>

I don't think performance is a that big problem for us. It's the unlimited
memory consumption that just keeps increasing. Virtuoso performs quite well
for most of our queries. I haven't gotten any complains from anyone about
Nepomuk in 4.10 being too slow. Whichever bottlenecks they may be, are
"most probably" in our code. Not theirs.

Even with the lowest memory settings (something I like running virtuoso
on), it slowly keeps going up until it takes over a GB at times.

* I have implemented one of those ground-up - from disk-access to the
> deductive/inferencing layer, so I kinda know what I'm talking about :)
>

Even though there are joins at every turn, I haven't found that to be a
problem with virtuoso. Though I have been dreaming about "maybe"
implementing something like this [1] on top of QtSql. While it has its
benefits, I'm not sure it would be a good idea. We should focus more on
features.

[1] http://cs-www.cs.yale.edu/homes/dna/abadirdf.pdf

-- 
Vishesh Handa
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/plasma-devel/attachments/20130217/542638ae/attachment.html>