[Foundation-l] Re: Development tasks and project needs

[Erik Moeller]

Michael-
> Do we have a conscious preference for having all sites on basically the
> same software configuration? For example, could we get something like
> dynamic listing, for which I can see the usefulness on Wikinews, without
> applying the same thing to Wikipedia where it might be detrimental? Is
> this sort of thing happening already more than I'm aware of? Why would
> this not be possible within the MediaWiki framework?

My take on this is that it is desirable to take a feature to a state where  
it is generally usable on all wikis, for multiple reasons:

1) If you give people a new tool, they may often use it in ways which you  
have not anticipated. For example, I can easily see how a dynamic list or  
RSS feed of articles matching a set of categories, sorted by date, could  
be used by
 - Wikipedia, for showing me the latest articles in a category I'm
   interested in, to systematically review them,
 - Wikimedia Commons, for subscribing to media feeds about certain topics
   (cf. Flickr),
 - Wikibooks, for getting new book modules directly into my RSS readers,
 - Wikiquote, for sending me new quotes about my favorite subjects,
and so forth.

There are areas in Wikipedia, such as the Wikipedia Signpost which you  
edit for, that would benefit from more Wikinews-like features. There may  
be people who want to maintain a blog on their Wikipedia user page. It  
pays off to think broadly instead of narrowly, to allow yourself to see  
more than one application for a particular technology.

2) The larger the divergence, the more difficult the Wikimedia Grid  
becomes to maintain. Our server admin volunteers often don't know what's  
going on in a particular project. It will be increasingly difficult for  
them to track down when something is wrong if every project has a  
different set of functions. If, however, something like Special:Feed  
exists as part of the general MediaWiki functionality, it is relatively  
easy for us to tell if it's causing problems.

3) It encourages laziness. The reason the dynamic list extension is not  
currently live is that it is not implemented in a scalable fashion. I can  
see the argument that it should go live on Wikinews anyway, but who knows  
that we have the proper replacement in place when we need it? It is  
undesirable to have all of Wikinews suddenly stop working because a  
developer decides that the extension no longer scales, and then have to  
code a replacement within a couple of days.

4) There is no particular reason why any of the features we discussed  
should not be able to scale.

5) The features we discussed are not fundamentally different from things  
such as some of our special page reports (most wanted articles), which we  
enable on all wikis. For all these features, we should look for ways to  
make them scale.

I think the MediaWiki developers are highly aware of the problem of  
feature creep. To extend our software in ways which go beyond its original  
scope is not automatically feature creep. The creepiness factor for me is  
actually higher with an approach where you build software out of lots of  
extensions which do not interoperate well and which are not maintained as  
part of the main tree. Take a look at TikiWiki to see what I mean.

Regards,

Erik