On Sun, Oct 12, 2008 at 1:45 PM, Charlotte Webb charlottethewebb@gmail.com wrote:
On 10/12/08, Gregory Maxwell gmaxwell@gmail.com wrote:
If you want almost everything linked then you want something that could be accomplished by machine. Better to run some userscript that autolinks dates/etc than to fill the article with low value links:
To me a "low value link" is one that pipes a noun into a verb, adjective, adverb, etc. e.g.
*[[malice|maliciously]] [[injury|injured]] by a [[baldness|bald]] [[table tennis|ping pong player]].
There are all kinds of low value links.
Your verb example is a sub-case of my own linking pet peeve: Linking some specific instance of something to a 'generic' article. In your example I'd expect "injured" to link to some article about this specific instance of injury, or at least the specific class of injury. It's not just noun->adfoo, it happens with perfectly noun-noun things as well. In your example I would have expected "ping poing player" to link to an article about the specific involved ping pong player.
Generally I see that as a pattern of linking to something boring and generic that I could easily look up if I were interested in it and in which reading about would not enhance my knowledge of the subject at hand very much.
I, and no doubt many other people, see many instances of date linking as examples of exactly the same class of bad linking: If I'm reading about the discovery of [[Uranus]] the material on "1781" and especially "March 13" are of no more value to me than "injury", "malice", or "baldness" are in your example. I do not intend to say completely without value, but not more than other generic things in the article text. Colors, common verbs, and other things I could easily search for if I were interested.
It seems that your valuation of links differs from mine, which shouldn't be a surprise: People can hold considered opinions which none-the-less differ.
Yeah, uh huh. *Sherman ordered them to [[March 2000]] furlongs eastward. *On channel [[12 June]] Cleaver is cooking supper. *Every [[December 30]] relatives show up uninvited. *On [[January 6]], [[1680]] homes lost electricity.
Those are all terrible sentences which would confuse or mislead readers even in the absence of links.
I have no doubt that you could cook up some examples which were were not ambiguous, but rejecting a 99.999% solution because it's not a 100% solution is a perfect way to abandon being good entirely.
There are two reasonable ways I see to close the gap to 100% there:
1) Have a 'grouping markup' in the text, or multiple link types. I.e. [[{12 June}]], it's not a link but it would indicate that the contained text is an atomic unit. This would assist all forms of machine parsing.
and/or
2) Instead of auto-linking text use JS to allow readers to highlight ANY contiguous span of text and click to get an "I'm feeling lucky" search result. This would allow article authors to link on the "you really should know about this" text, but still provide readers with the ability to explore generic concepts with which they may lack familiarity.
(as a side benefit the same linking JS could be provided to third party websites which might find it useful and which would drive more traffic to Wikipedia)
while the manual linking allows the identification of the significant, a task the machine is ill suited for, and preserves the ability for people who do not care for everything being linked to still browse with the most significant links provided.
Frankly I don't trust the community to decide which links are significant.
Then you're wasting your time arguing about this, aren't you? There are millions of pages on Wikipedia, so you couldn't maintain the linking personally even if the community would let you be supreme-god-of-linking. If the community can't achieve your definition of success then your definition of success simply can't be met. Our attention would be better spent on achievable goals, no matter how good we decide your criteria to be, and the mass linking of dates is by no means a universally agreed good thing.