Implicit Link Names

[Jason Clark]

On Apr 2, 2004, at 2:38 PM, european bob wrote:

> I don't think the risk is negligible, although I'm in favour of
> [implicit links]. For example, someone cited the example of links that
> look like this[1]. They are extremely common in e-mail, and Markdown
> will be making them look odd. (Actually, I even disagree they're links 
> -
> I always understood them to be footnotes ;)

I'm still don't see a problem... I agree with you that such links are 
footnotes.  Here's the example in question (posted by Aaron Swartz):

	[Joe][1] [said][2] this [the other day][3]:
	
	> Can you believe this guy? He can't get through a single paragraph 
without lying:
	>
	>> I'm proud to have been chosen by the American people. [1] I feel 
good about this country's direction. I've created jobs. [2] I've made 
the environment safer. [2] And I've stuck to my word -- I did just what 
I promised I'd do on the campaign trail. [3] I'm a guy you can trust.
	
	[1]: http://joe.org/
	[2]: http://joe.org/weblog/29828
	[3]: http://me.org/calendar/2004/04/23

The problem is that the author used linknames like [1] [2] [3], and 
then blockquoted some text that appears to use email-style footnotes.  
It seems simple enough to write the original block like this:
	[Joe][x1] [said][x2] this [the other day][x3]

and fix the links accordingly, if you are going to be quoting text with 
email-style footnotes.  It's even easier to write this:
	[Joe] [said] this [the other day][day]

	[Joe]: http://example.com/1
	[said]: http://example.com/2
	[day]: http://example.com/3

If you actually want to have email-style footnotes instead of links, 
just escape the link:
	This is a test[1]

  	\[1]: http://example.com/footnote

Remember, if you don't define a link, it doesn't become a link.  I 
thought the whole concern was over accidentally triggering a link.

> I think the problem is basically what happens when you have a link 
> which
> doesn't have a definition. Currently, that falls back to a explicit
> bracketed thing. That strikes me as difficult, because it means a
> spelling mistake will stop it working and - worse - there is no way you
> can detect it at all.
>
I must not be understanding your point.  Let's try an example:

	This is a [link], and so is [thsi]
	
	[link]: http://example.com/1
	[this]: http://example.com/2

I made a typo on the second implicit link.  MD should produce this:
<p>This is a <a href="http://example.com/1">link</a>, and so is 
[thsi]</p>

The (intended) second link is broken, but I see it when I view the 
output (thanks to the [] becoming visible).  No different than this:

	This is a [link][link], and so is [this][thsi]
	
	[link]: http://example.com/1
	[this]: http://example.com/2

>
> If there were cruft to get rid of, I think I'd prefer the inline syntax
> to go ;) I don't think you can get rid of [link][id], because you may
> always want a nickname (link text could be very long). Simply outlawing
> an empty id seems a bit pointless.
>
I wouldn't suggest getting rid of [link][id].  I call [link][] cruft in 
this context because it would produce precisely the same output as 
[link] so why support two syntaxes, especially when one has extra 
punctuation and causes more visible interuption to the text?

> I think I'd prefer something like [inline implicit link]* and
> [inline implicit image]! though... I'm still slightly uneasy about just
> plain [link].
>
I'm really not trying to be argumentative, I just still don't see how 
[this] is likely to cause unintended results... at least not often 
enough to outweigh the potential benefits.  If John releases a version 
to the list that allows [implicit links], perhaps we'll find examples 
in use that we won't think of otherwise.

Jason Clark <jason at jclark.org>
http://jclark.org/weblog/