Which Wiki Would Work?

January 22, 2009 9:01 AM

I want to start a wiki. Help me decide how best to go about it with long-term planning and goals in mind.

I have an idea for a Wiki that has not yet been created (is it possible?? Yes it is...). It is related to a topic where I already have a web presence with over 100k unique visitors per month to my domain so I believe I have a core base of people who will jump into this Wiki and help make it a success.

The end goal for the Wiki would be something that would have over 100,000 articles, and likely be somewhat picture heavy with between 1 and 50 pictures per article.

But to start, I cannot decide between using my current webhosting situation using a MediaWiki install or using wikia.com to set up the site.

Currently I own several of my own domains, and I have them hosted with Dreamhost (which has given me some fits in the past but {knock on wood} has been great of late). I have their "unlimited" storage and bandwidth option.

My vision is fast initial growth eventually tapering off to regular updates, and huge growth over time.

By the same token, I would like to be able to do affiliate links (Amazon, etc.) and try to get some advertising revenue for my work if it does take off, and I also want to leverage the internet presence I have currently, my "brand" if you will.

So does anyone have experience with Media Wiki or Wikia? Any preference between them? If I start with one and need to transition to another, is that do-able? How much work involved, etc.

Thanks!

posted by arniec to computers & internet (10 answers total)

I used to have an (internal) mediawiki site and found it very cumbersome for image attachments.
posted by devbrain at 9:15 AM on January 22, 2009

You can do a one-click install of MediaWiki with Dreamhost, so why not try it out? You can have an empty wiki up and running in minutes to play with. I've done it myself with no problems. Regarding image upload, I've found the MediaWiki extension SpecialUploadLocal to be extremely useful. You can FTP a bunch of images to your server and then use the "Upload Local Files" page to get them into the wiki.
posted by Otis at 9:29 AM on January 22, 2009

I installed MediaWiki on one of my servers just to play around with and found it incredibly, painfully slow compared to Wikipedia or any of the WikiMedia installations, but I haven't looked into that and I wonder if I just need to do some PHP optimizations or something. (On the other hand, given the sheer amount of stuff MW must do internally, it wouldn't really surprise me if it's very resource-intensive.)

But I think that if I was going to launch a wiki I'd want it on my own server, for the sake of having greater control over things like advertising and also because I'd probably want to be able to install any MW extensions I desired (which I'm not sure if Wikia allows.)

devbrain - on my copy of MW, although it was slow, uploading images worked the same way it does on Wikipedia and elsewhere. Do you mean it was cumbersome in general the way it is on those public sites (which I'd basically agree with) or are you describing something different?
posted by XMLicious at 9:35 AM on January 22, 2009

Its tough to answer this without knowing how much traffic you'll be generating. Each wikimedia load polls that database. Its not a flat file system, so you'll need hardware that can handle this. If you expect moderate to high levels of traffic then I would absolutely not go with dreamhost or any cheap hosting solution. I would recommend buying your own server (or two) and colocating them.
posted by damn dirty ape at 10:50 AM on January 22, 2009

I'm with XMLicious: if I were you, I'd feel better about hosting my own MediaWiki than relying on Wikia. And with Dreamhost, you'll always have the option to painlessly move to a dedicated server if your future load demands it.
posted by paulg at 12:31 PM on January 22, 2009

Also, this PDF presentation on Wikimedia architecture and this Wikimedia servers page show some of the things they're doing to bear the Wikipedia load. It is a little fancier than your run-of-the-mill LAMP installation.
posted by paulg at 12:37 PM on January 22, 2009

Each wikimedia load polls that database. Its not a flat file system, so you'll need hardware that can handle this.

I agree with you dda that it appears as though MW itself probably requires lots of horsepower which arniec should consider. But I want to mention that this isn't a general problem with database-backed applications; for example, on the same server that MW runs slow for me on, Drupal runs just fine as do a variety of other PHP and Java database-backed applications.

From experience as a user of Wikipedia and seeing the sort of features that MW is implementing I'd guess it's probably that MW is simply doing a lot of stuff. (Though of course it's also possible it's spaghetti code and that's why it's slow, I haven't examined the internals that closely.)
posted by XMLicious at 2:50 PM on January 22, 2009

There are two beliefs I've gathered from my attempts to use Wikis:
- Non-technical people basically can not or will not use non-WYSIWIG wikis. So like if your userbase isn't all nerds, MediaWiki is out. Even fairly technical people really don't like using it.
- Wikis are extremely difficult to do right and depend on weird community dynamics. One thing is wiki gardening: it quickly becomes a surprisingly big chore that no one wants to do, and then people stop contributing to the wiki cause they don't know where to put stuff. Even relatively tight, highly technical, incentivized people generally fail at this. Look at the Rails wiki over the last few years, for example. It seems there are two kinds of successful wikis, but on the public internet, they are both fairly rare:
- Sui generis wikis from whom its hard to draw replicable lessons. Basically there are two of these to my knowledge: wikipedia and the c2.com wiki.
- Wikis made by obsessive, nerdy fan communities, like the Star Trek wiki. Star trek fans are given to crazy obsessive detailing and documentation of star trek, so they will naturally build and take care of a wiki, but like...there's tons of fans of electric guitar amps, and I've never seen a good amp wiki. Even in communities that have these sort of obsessive, detail-oriented fans, when they lack technical leanings, they don't seem to generate good wikis. It seems that traditional websites made by a small group of people generally beat wikis. That is, like most open source projects, the level of community involvement is quite low. A lot of those "museum" sites are like that.

I guess the advice I'm trying to give you that is relevant to your question is, for planning purposes, if your target users are not nerdbombers, use some sort of wiki-like software that has wysiwig tools, not mediawiki/instiki/moinmoin, etc. Second, consider if a wiki is really likely to be successfully built by these users before you invest too much time it. There are really very few good wikis out there. Googling your name reveals some star wars links, and your profile shows you've used some technology tags, so perhaps your audience is good.
posted by jeb at 3:44 PM on January 22, 2009

I don't know of any visual editor that fully supports wiki code but adding an HTML WYSIWYG editor to a web application is dead simple if you know any javascript, so they're available in pretty much every modern web application. In MW it's just a matter of selecting and installing an extension. (I assume that the reason none of them have been installed on Wikipedia et. al. is the full-support-of-wiki-code thing.)
posted by XMLicious at 5:20 PM on January 22, 2009

But I want to mention that this isn't a general problem with database-backed applications

I agree. I guess my point is that this isnt some lightweight application that's heavily cached like a simple html page served up by apache. Its pretty hefty. I run an in-house wikimedia installation at work and its surprisingly slow on dedicated modern hardware. I cant imagine the nightmare it would be on dreamhost. Their servers always seem pretty beaten up.

Its also worth nothing that people only have so much patience. I think HCI guidelines are something like 4-7 seconds max to load a page after clicking. If he starts his project on shared hosting and these pages start taking longer than 6 or so seconds then his potential users will just quit before the wiki can even get started. You gotta get good performance from day one, especially if you are trying to attract people to give you their time and effort for your content.
posted by damn dirty ape at 7:33 PM on January 22, 2009