Re: [kDev] MediaWiki developers evaluate Kendra Base...

[Daniel Harris <daniel@xxxxxxxxxxxxx>]

Hi Erik and All,

On Oct 16, 2004, at 5:11 am, Erik Moeller wrote:
> Hi Daniel,
>
> thanks for your mail, and for adding the Kendra mock-up to the 
> Wikidata page on Meta. What I wasn't able to figure out: Is it 
> possible to see the revision history of an individual property?

We have a chunk more coding to do before we hit our goal for this stage:

http://www.kendra.org.uk/wiki/wiki.pl?KendraBaseWishList

Revision history is in there.

> Are diffs etc. implemented?

I would expect so. Neil Harris has been writing all the Python code so
far and I don't know Python myself. You can see all the code here:

http://www.kendra.org.uk/develop-code-kendrabase.php

> I'm also not quite sure what's possible in terms of data relations, 
> though the simplicity of backlinks has some beauty to it.
>
> I think in general KendraBase's goals are very similar, if not 
> identical, to those of Wikidata.

I'm not so sure and correct me if I am wrong:

Wikidata aims to "centrally store and manage data from all Wikimedia
projects". Whereas Kendra Base aims to be a distributed platform
enabling disparate databases, marked up in different
languages/terminologies to interoperate as if they were one centralised
database from a user's viewpoint.

So perhaps we have the same goal for the experience of the
viewer/searcher of the information. But in terms of the publisher I
think we are quite divergent.

> There are quite a few reasons why the KendraBase implementation 
> cannot be the base of the Wikidata project, however.
>
> Chief among them are:
>
> * MediaWiki. I consider it the best wiki engine that is currently 
> available, with the possible exception of the proprietary Jot (which 
> has Kendra-like features). It is also the basis of all Wikimedia 
> projects. To use an entirely different software would be undesirable,
>  because it would be incredibly hard to provide a consistent user 
> experience, and because we would thereby abandon the goal to 
> integrate the idea of wikidata into any MediaWiki-based wiki.

That fair enough.

> To clarify the last point, while Wikidata.org, in my vision, will be 
> a separate project, we will want to use Wikidata functionality in 
> virtually all Wikimedia projects. For example, it would be desirable 
> to have structured, searchable data about images stored on the 
> various wikis (caption, copyright status, photographer, etc.)

Exactly.

> * Scalability. As I understand it, you are using UsemodWiki. While it
>  is very nice, Wikipedia.org alone is one of the top 300 websites 
> already and growing constantly. MediaWiki has been prodded by dozens 
> of developers to get decent caching on all levels: filesystem, 
> database, shared memory, Squid proxies, user client. It has been 
> systematically vetted for slow queries that could kill the server. It
>  is tried and tested in a very harsh environment. Mind you, Wikipedia
>  originally started with UsemodWiki, but it's simply not up to the 
> challenge of running a site of this size.

That's why I glad we are having this conversation. Because your
evaluator didn't get all the facts right. We have a normal wiki where we
do wiki things and, yes, we use Usemod.

However, Kendra Base is an entirely separate application and is using an
entirely different language. It has nothing to do with any other wiki
software. All completely home grown Python code.

Also, Kendra Base is definitely not a wiki with database functionality
bolted on. It's a flexible datastore with a forms interface to
input/search information about objects. Granted, the forms that we are
using look like a wiki because we like the simplicity of the interface
but we will also be able to edit objects entirely by other types of
forms with drop down menus, etc much like you show in your mock up.

> * Usability. Much as I love KendraBase's feature set, if I had 
> discovered it a few months earlier, I probably wouldn't have spent 
> much time with it - simply because the user experience is quite 
> confusing. It is unexpected to see programming terms like "instance",
>  "class", "ontology" used in the context of wiki editing, and the 
> navigation through the system is not very clear.

What words/terminology would you like see used to describe these
features? We could do with your understanding here.

> I realize it is currently in the proof of concept stage, and at that 
> it succeeds marvelously. But I think turning what is currently there 
> into a base that would pass the usability test of exposing it to the 
> teeming Wikimedia masses would require almost as much effort as 
> writing something new from scratch. Good user interfaces are really, 
> really hard, and that aspect of software design is often 
> underestimated.

That's fair enough comment. So, please, how would you design the
interface? You will need to do this for Wikidata. Do you have any ideas yet?

> With this in mind, I will definitely be spending more time trying out
>  KendraBase and observing its progress. I'm sure there is a lot we 
> can learn from it for Wikidata development.

And vice-versa! ;-)

> I'm currently occupied getting another revolutionary project off the 
> ground (Wikinews), but I should have some time to devote to this in 
> November.

Nice one.

> Now, I'm not sure what your long term plans for Kendra are. If you 
> see a way this project could be merged into the larger Wikimedia 
> effort, I would definitely be interested in discussing that.

Well, that's an interesting idea and I'm sure we will look at that very
seriously as we continue development of Kendra Base.

> While the Wikimedia Foundation is not yet the best-funded 
> organization (we just raised $50K in two weeks, mostly to cover 
> hardware costs), I believe it is quite likely that we will get 
> substantial grants in the near future.

Ditto! ;-)

> A requirement for such a cooperation would be that any future work on
>  this project is done on the basis of the MediaWiki code in 
> collaboration with the existing MediaWiki development team.

The reason for creating Kendra Base is to provide Kendra Initiative with
a collaborative design tool to enable us to build an open marketplace
for digital media. For more information see:

http://www.kendra.org.uk/wiki/wiki.pl?Kendra_Initiative_Executive_Summary

Kendra Base is not a standalone project but a small part of a much 
larger one.

Kendra Initiative's aim crosses the borders of "free" media and
"commercial" media. We aim to provide for both. As it's all really the
same stuff. Just bits of information with permissions attached.

> Again, whether this is desirable to you depends entirely on your 
> goals. Wikimedia is the organization which is in the best position to
>  get Wikidata off the ground - it has a huge community of users who 
> are just waiting for something like this. So if your goal is to get 
> ontology tools into the hands of the masses, I think a partnership 
> with Wikimedia would definitely be the way forward.

I definitely think you are right. Partnership would be good. For the
next few months I think it best to continue to the end of our current
development stage. In parallel with this I think we should start to
flesh out how a partnership would work. Yes?

Let me know if you want to get our newsletter (one every few weeks, with
a following wind). Or register here:

http://www.kendra.org.uk/register.php

Cheers Daniel