[Nepomuk] Re: Gsoc Proposal draft

[Sebastian Trüg]

Hi Swair,

good to hear that you are interested. A few comments on your proposal:

In general: When it comes to project management the idea is not to use
tags (as in string keywords) but to use actual pimo:Project resources.
This means that each project is a fully qualified RDF resource which has
a label, a description, and any other properties one might want to
attach to it (like start date). Then other resources like files or
webpages are related to that project.
(To be precise this is not that different from tags in nepomuk since
they are also RDF resources - simply with a different type, namely nao:Tag).
More below:

On 03/09/2011 03:23 PM, swair shah wrote:
> *Proposal Title:* Project Integration 
> 
> *Motivation for the proposal:* Before Project Integration Gsoc project,
> I had started doing some research on another Nepo muk project : Building
> a "real" query parser, and started reading webpages for writing context
> free grammar to parse queries. Even before starting anything related to
> coding I realized the need of a "Project Integration" which would keep
> all the previously visited webpages, local files associated with the
> same project, even the mail conversations I would have regarding the
> project and at the same time the contacts related to the project. 
> 
> I believe that Nepomuk is a very powerful project and its not being used
> by KDE users to its full potential, probably because the lack 'front
> end' applications like the Project Integration tool which would really
> show Nepomuk in action. 
> 
> Project Integration can be really powerful for development projects by
> semantically linking code repositories, reference docs, tutorials,
> videos, contacts etc. with a particular project. Apart from development
> projects, such a framework can also be used for creating and searching a
> semantic database for non development projects.
> 
> *Implementation Details:  *
> About webpages:
> There should be a way to tag webpages to the project db. Can use
> annotation in Konqureror. For other browsers a currently running webpage
> can be tagged with the project. (krunner can be used like say,
> projectname:Tag:webpage)

Webpages can already be tagged from Konqueror via a context menu. The
code can be found in the nepomukannotation project on projects.kde.org.
However, as said above one cannot relate a webpage to a project. I
envision this to be very simple, even combined with something like a
current project. Again we already have a service maintaining a general
current resource. This could be combined.

> Files:
> Any files can be tagged with the project name. There can be support for
> automatic tagging so that user don't have to go and tag everything
> manually (may be files edited in one 'session' can be included in the
> same project) but i believe manual tagging is better as user would have
> full control over the files and pages that are included in the project
> metadata.
>
> Other contributors:
> If there are other people working on the same project, their metadata
> will be connected with my metadata. Email converstations with the
> coloborators which have subject/content which contains name of the
> project will be added in the rdf store for the project.

There is only one big RDF store - Nepomuk. Everything is stored in there.

> *Tentative Timeline:
> 
> *First half of may: Go through nepomuk code. Learn how to use virtuoso.
> Start working on some basic RDF triple store creation to get experience.

I don't really get this. Nepomuk has an RDF store and there is no need
to create a new one.

> Mid may- Mid june: Implement the support for local file tagging and
> start working on webpage tagging for a particular projects.

"local file tagging" sounds like what can already be done today.

> Mid June - Mid July: Implement support for metadata sharing with
> different contributors woking on the same project. 

This is very ambitious and needs a lot more detail.

> Mid-July - August Begining: Create a GUI widget which can be used to
> search project content based on either project name or a file name,
> contributor name who is associated with the project. It would return the
> list of all the file, webpages, users.

All in all your proposal is way too short. It is not clear what you
intend to do, especially for people that have no idea what is meant by
"project integration" in Nepomuk.

Cheers,
Sebastian