Search Mailing List Archives


Limit search to: Subject & Body Subject Author
Sort by: Reverse Sort
Limit to: All This Week Last Week This Month Last Month
Select Date Range     through    

[protege-discussion] collab protege - status and roadmap?

Tania Tudorache tudorache at stanford.edu
Mon Dec 3 13:46:38 PST 2007


Hi Jens,

Thanks a lot for the feedback.

usenet at mnet-mail.de wrote:
> Hi,
> as one of the early users I played with collab protege and I think it's a great
> idea!
>
> In an earlier thread this year I posted my experience. There are some problems
> with collab protege which meanwhile forced me to not use it any more on a
> production system. 
>
> One big issue for me was saving of the annotation ontology. On a multiuser,
> database based protege installation, saving annotation ontology to a file
> re-introduces all the problems coming with that.
>
> I'm using automatic interfaces to get data into the ontology, which can lead
> to a rather large annotation ontology. Saving the ontology often failed and we
> ended up with a zero byte annotation ontology file loosing all the history.
> Performance with collab protege is - as exptected - not good compared to plain
> protege.
>
> So what are the plans for the future?
>
> For me, the biggest step would be removing the file based annotation ontology
> and enable saving as a database project. I tried to convert annotation onto to
> database backend, but it does not work.
>   

Using the database backend for the annotations project is one of the 
items that we are going to implement very soon. Very likely this year or 
early next year.

I would like to know more about the performance problems. When did you 
experience the delay? Browsing classes or instances? What was the 
average console ping time between a client and the server?

The problem with the annotations project is that the access to it is 
random. With the main ontology, the server can predict what frames the 
client will need and can precache them. For example, if a user clicked 
on a class, very likely he will open that class, so the server will 
precache the subclasses and other related information. In this way, when 
the client will ask for the subclasses, the server can provide them very 
quickly. The access to the annotations project is random, in the sense 
that it is not so easy to predict what are the next needed annotations 
by the client. So, we need a different caching strategy on the server. 
We already have one implementation, which needs tuning for different 
working conditions.

Thanks again for the feedback,
Tania



More information about the protege-discussion mailing list