Wikipedia: Historical Wikipedia Pages/ArbitratryLinkingDiscussion

The other day, on a page I cannot find now, someone suggested that we might switch to a version of Wiki which supports arbitrary linking. The idea is that, particularly for an encyclopedia, having arbitrary links is better than the amusing but odd wiki capitalization conventions.

It strikes me that a very simple system would be to be able to link like this: <<< Albania >>>.

Is there a wiki software that can support this? -- Jimbo Wales

["Free linking" like word was later added to UseModWiki. See below for discussion of that feature.]

I just had another thought. Perhaps arbitrary linking using <<< >>> should be limited to single words only. For example, a list of United States might be written like this:

<<< Alaska >>>
NorthDakota?
<<< Wisconsin >>>

That is, if you use <<< >>>, the rule is that there has to be only one word inside there.

The whole point here is to get away from weird names for single-word articles, things like NuPedia or AlaSka?. Is there a wiki that can do that? Or, should I just add the capability to this one? Or, is this whole idea really bad for some reason?

Adding word or <<<word>>> -style links to this wiki should not be very difficult. The main disadvantage of such links is that they would not be automatic. That is, if someone typed the plain word "Alabama", it would not become a link automatically. (The user would have to type something like "Alabama".) On the other hand, I'm not sure that every instance of simple words should be a link. Requiring explicit links might make the page look cleaner.

If this wiki community really wants more free-form links, I'd be willing to take a look at adding word-style links as an option in UseModWiki. --CliffordAdams [This feature was added in February 2001.]

This issue contains a conflict in the different uses of the system. On one hand we can use a simple enough syntax to mark the links like the funny capitalisation of letters and on the other hand the writer of each entry will have to put extra thought into what is a link and what is not.

I think that since the ambition is to build an encyklopedia we should go for the latter. An important thing with an encyklopedia is the words and their spelling and the funny capitalisation destroys the charm of the language. In some extreme cases there might even be a difference in the meanings because of the capitalisation. (Example from the top of my head: March n. month, march v. to walk). I would also like the possibility to write names with their correct spelling including the spaces like George W. Bush and North Dakota.

Actually I would like to make it even more complicated for the writer. I would like the Wikipedia to provide a possibility to link to the correct version of a word if words have several meanings like this: York (linked to York/PA) is named after the English city (York linked to York/England?). and have this appear as "York (clickable) is named after the English city York (clickable)".

Since this makes it harder on the person writing the texts we need tools to help him. I have identified two important set of tools:

First a tool that during preview goes through the text and signal all words that are present in the encyklopedia and thus are immediate candidates for linking. This is good as a help in identifing words or word combinations that are present elsewhere. The tool could also help him in choosing the correct version of the word.
The second tool that is a tool that when we add a new word, goes through the whole wiki and finds all places where this word or combination of words is used without a link. Kind of like the search function except that it does not include the words linked here (or linked to some other entry in the case where we have several entries with the same name).

A third more elaborate tools would be for those that want to spend time fixing all the errors that the persons not running the second tool would insert into the database and that would be a tool that finds (at random or in some order) entries with words or word combinations that are now defined but not linked to.

I realise that my technically ordering and systematising mind is what makes me wish this but I can't get out of it. --LinusTolke

: Multi-word links with spaces would require some more effort than single-word links. I would really like to implement a system which allows nearly arbitrary text in a link, even including characters like / or ! which may not be allowed in filenames. On the other hand, I could quickly add something like TWiki's [Some Phrase Here]? links, which would link to a page SomePageHere?.

: I agree that search tools to find potential links would be very useful. One possibility would be to make a list of pages in date order (newest first), with all the potential links listed next to the name of the page. Editors could then watch the most recent edits for potential links. If this wiki moves in that direction, I'm willing to work on better search tools. --CliffordAdams

[Links like word are now an option in UseModWiki 0.91. (Hopefully Wikipedia will upgrade soon.)]

I think it wouldn't be hard to write a conversion-helper script to semi-automatically fix the current strange capitalizations. The script would first look for all wikinames and create one large list. A human editor would go through the list and delete any of the "good" multi-word names (like PopularMusic), which would end up with a list of "strange" wiki-names (like PolanD or TopOlogy). The second part of the script would take take the edited list of one-word names and replace them with word links. Those one-word names which already have defined pages could also be moved/changed automatically, along with the RecentChanges database file.

The whole conversion would probably take only a couple hours, much of which would be editing the list. I'm willing to write the conversion scripts and even do the editing if the maintainers of Wikipedia approve. --CliffordAdams

(I think this sounds great -- Jimbo Wales)

This sounds good to me, but Jimbo Wales is the Wikipedia programmer. Feature request: make it possible for a link written [[like this Jimbo Wales]] to be displayed with just the words "like this" linked, and pointing to the "Jimbo Wales" page. However you do it, it would be dandy to be able to link to a page without using the page name. On the other hand...maybe we don't want to make that possible! -- Larry Sanger

: I've implemented a simpler syntax like like this, which is closer to the regular URL format: [URL description]. Trying to parse nested pairs of brackets would get pretty ugly coding-wise.

: For spaces in links, they are translated into _ (underline) characters in the URL, and translated back into spaces when displaying the link. This allows you to type Jimbo Wales, which would be displayed like "Jimbo Wales" (as a link). If you wanted to link to the page (from outside the wiki) you would use a nice URL with an underline (like http://www.wikipedia.com/wiki/Jimbo_Wales), rather than an ugly one with a + or %20 for each space character.

: Hmmm... If one handles spaces like that, perhaps this wiki could completely dispense with the strange capitalization rules and just use [word or phrase]?-style links. There would be less "accidental linking", but this wouldn't necessarily be a bad thing. (I'd certainly want to create link-helping tools, however, to find appropriate links.) --CliffordAdams (soon to be [Clifford Adams]??)

I love weird capitalization! I think we should not be too prompt putting a straight jacket on Wikipedia --OprgaG

I have a suggestion that might give us the best of several worlds. Continue with multiple capitalization as the main tool used by authors, but allow one to assign single word synonyms with no capitalization so that they can be recognized for indexing when they appear in any article. --DickBeldin

: I'm not sure I understand what you're asking for here. I don't think it would be a good idea to automatically link single words every time they appear--it could lead to far too many links on pages. One can already create synonyms with the redirect command--see MozarT for an example (redirected to the composer's full name). --CliffordAdams

I don't understand why you want spaces to be converted to and from _ (underline). Isn't the space character good enough as it is? Is there some conceptual meaning of the space that I don't understand? --LinusTolke

: Briefly, spaces in UniformResourceLocators? (URLs) are forbidden in the UniformResourceIdentifier? (URI) specification. (URLs are a subset of URIs.) From RFC 2396: "The space character is excluded because significant spaces may disappear and insignificant spaces may be introduced when URI are transcribed or typeset or subjected to the treatment of word-processing programs. Whitespace is also used to delimit URI in many contexts."

: For URLs, the valid ways to represent a space are the "+" character (within a "query component") and the "%20" escape sequence. Both look relatively ugly compared to an underline: compare "Jimbo_Wales" with "Jimbo+Wales" or "Jimbo%20Wales". Internally my code also uses the _ character for filenames, as some utilities on some systems do not deal well with spaces in filenames. However, aside from the URL issue, you should not notice the difference. --CliffordAdams

Yeah, I think you're right. I think the freeform linking may be the thing you need. Most of the cruft comes from the poor choice of a LinkPattern? (suited best to http://c2.com/cgi/wiki?PatternLanguages). -- SunirShah