...or your users will!
"Doing Terrible Things To Your Code http://blog.codinghorror.com/doing-terrible-things-to-your-code/" is a good read on testing by Jeff Atwood on his blog, Coding Horror http://blog.codinghorror.com/. I also found the "falsehood" snippets poignant—maybe we should come up with some for Wikipedia ;-). Here are a couple off the top of my head, at least for "official" Wikipedia instances:
1. Wikipedia sites all have standard ISO/BFC prefixes (see sitematrix https://www.mediawiki.org/wiki/Special:SiteMatrix) 2. A site's main page is always titled "Main Page" (also see sitematrix) 3. A page's revision is a reliable snapshot of its content (nope: transclusions [and images?] can change independent of a page revision) 4. API error messages are plain text (nope, can contain HTML https://phabricator.wikimedia.org/T107082)
Interested in hearing "falsehoods" you've encountered.
Cheers,
Brian
Regarding your 3rd point, we have discussed this in our organization (NASA) and we have considered some ideas on how to improve the situation. Between changes in templates and other transcluded material including content re-used via Labeled Section Transclusion and query results via Semantic MediaWiki there are many ways a page could change without "direct" documentation on the page itself. As we are considering how to share data between a network/farm of wikis, this compounds the problem. Just to throw an idea out there, we talked about possibly incapsulating any transcluded material in a very thin box with some form of link to the source. This might work for "bulk" material like infoboxes entirely made up by templates and queried data. It might also work for paragraphs transcluded via LST. But it probably becomes a nightmare for bits of data like a single-digit value for the mass of an object. Maybe we could instead make transcluded content a different color or font style with a hover-link to the source.
I'll add that this isn't just a point about being able to pull up an historical version of the page (without portions of it being the current version due to transclusion), but it's also a point about confusing non-"power users". I have encountered many users who ask me "How do I edit this value on the ABC page?" when they don't see it in the edit source view. Sure, I can educate them on the different mechanisms we use, but it is a very common question. It seems we are straying from the core characteristics of a wiki[1] (easy to use, promote associations, invoke sustained usership).
I'll add in a 5th point straight from the blog: People have one name. This isn't typically a problem on open/public MW instances where users are used to the optional anonymity of an online username, but in firewalled/corporate/enterprise environments the users might not go by their username. John Smith in accounting would be difficult to mention via Echo/Thanks if you didn't know his username was jsmith5.
Here's a few more thoughts on falsehoods:
- Once a user joins a wiki, they will continue to contribute (general decline of usership) - If a wiki has lots of users, the content is reliable [2] - If a user watches a page, they will review revisions [3] - If you pay people to use a wiki, they will [4]
[1] https://en.wikipedia.org/wiki/Wiki#Characteristics [2] http://enterprisemediawiki.github.io/slides/WikiAccountability2/#/2
The content isn't very reliable if only a handful of people are reviewing it.
[3] http://enterprisemediawiki.github.io/slides/WikiAccountability2/#/4
In [3], this is a network graph of the state of the EVA Wiki when we realized users don't watch pages and review revisions. Orange dots represent people, blue dots are pages. Clusters of blue dots are groups of pages watched by one person. Red lines indicate pages with revisions not yet reviewed by watchers.
[4] http://enterprisemediawiki.github.io/slides/WikiAccountability2/#/20/2
We are still learning how to employ intrinsic motivation because extrinsic motivation cannot sustain continued usership.
Daren
On Thu, Jul 30, 2015 at 9:35 AM, Brian Gerstle bgerstle@wikimedia.org wrote:
...or your users will!
"Doing Terrible Things To Your Code http://blog.codinghorror.com/doing-terrible-things-to-your-code/" is a good read on testing by Jeff Atwood on his blog, Coding Horror http://blog.codinghorror.com/. I also found the "falsehood" snippets poignant—maybe we should come up with some for Wikipedia ;-). Here are a couple off the top of my head, at least for "official" Wikipedia instances:
- Wikipedia sites all have standard ISO/BFC prefixes (see sitematrix
https://www.mediawiki.org/wiki/Special:SiteMatrix) 2. A site's main page is always titled "Main Page" (also see sitematrix) 3. A page's revision is a reliable snapshot of its content (nope: transclusions [and images?] can change independent of a page revision) 4. API error messages are plain text (nope, can contain HTML https://phabricator.wikimedia.org/T107082)
Interested in hearing "falsehoods" you've encountered.
Cheers,
Brian
-- EN Wikipedia user page: https://en.wikipedia.org/wiki/User:Brian.gerstle IRC: bgerstle _______________________________________________ Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
1. You can always assume that all your input is magically converted to UTF-8 NFC (Not if you're running from a maintenance script) 2. You can do things like call the API (internally) to edit a page when a method from $wgParser->parse() is on the stack (Well you can, but the page you're parsing is going to be "uniq" if you get my meaning)
- A page's revision is a reliable snapshot of its content (nope:
transclusions [and images?] can change independent of a page revision)
Since you put a question mark here, to be clear - image versions can change independent of revision text. So can other things, like parser functions, magic variables ({{NUMBEROFARTICLES}}) or tag extensions (e.g. <CategoryTree>). Not to mention site css/js can make things look totally different. And then its a whole other confusing mess if FlaggedRevs is installed.
-- bawolff
On 7/30/15, Brian Gerstle bgerstle@wikimedia.org wrote:
...or your users will!
"Doing Terrible Things To Your Code http://blog.codinghorror.com/doing-terrible-things-to-your-code/" is a good read on testing by Jeff Atwood on his blog, Coding Horror http://blog.codinghorror.com/. I also found the "falsehood" snippets poignant—maybe we should come up with some for Wikipedia ;-). Here are a couple off the top of my head, at least for "official" Wikipedia instances:
- Wikipedia sites all have standard ISO/BFC prefixes (see sitematrix
https://www.mediawiki.org/wiki/Special:SiteMatrix) 2. A site's main page is always titled "Main Page" (also see sitematrix) 3. A page's revision is a reliable snapshot of its content (nope: transclusions [and images?] can change independent of a page revision) 4. API error messages are plain text (nope, can contain HTML https://phabricator.wikimedia.org/T107082)
Interested in hearing "falsehoods" you've encountered.
Cheers,
Brian
-- EN Wikipedia user page: https://en.wikipedia.org/wiki/User:Brian.gerstle IRC: bgerstle _______________________________________________ Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
wikitech-l@lists.wikimedia.org