The first thing I noticed was that he was making the common claim about things which were reader-optimized : that the writing would later be optimized by better tools. Because I wanted to respond to that, it was the most natural thing in the world to link to the paragraph where he said it. Using its purple number.
But then a much bigger thought struck me.
Purple numbers totally make sense in the context of a blog.
Because when I'm making links to a blog entry, I'm typically already looking at it on the screen in front of me.
Grabbing the purple number attached to the paragraph is as natural as grabbing the url of a page, or a permalink to an entry. And indeed there's a continuity between the fine-grained permalink of blog posts and the finer-grained purple-numbered paragraph. The action and logic of linking is the same.
My initial discussion with Chris, was, in contrast, about purple-numbers in wiki, which is what E.E.C. and Blue Oxen have always appeared to me to stand for.
Now, the reason I felt a disconnect was that my typical link-making behaviour in my wiki is different from my link-making behaviour in blogs.
In ThoughtStorms, what usually happens is I get or find an idea to write about, I think up a suitable name for it, and create a page about it. As I'm writing this page, all kinds of associations with other ideas pop into my mind.
Some of those ideas I know have pages about them already, and I remember well what the name of that page is. Some of those ideas I know probably have pages but I don't know for sure what the name is. Some, I know don't have a page, but should. In the first case, I put in the name I know. In the second, I guess the most likely name. In the third, I think of the first sensible name that seems to get a handle on the idea.
In other words, I'm essentially doing something rather like tagging. I'm conjecturing how to reach other pages about an idea, based on my own ordinary language assumptions.
After I save the page, I sometimes find out that my guess was wrong. A page I thought existed under a particular name, didn't, in fact, exist.
At this point I might do one of two things.
I might realize the name was just wrong. And re-edit the page to fix it. Or I might decide that the name I conjectured was another worthy way to get at the idea; and so go to that non-existant page and turn it into a #REDIRECT to the real page. (In other words, I define it as a synonym.)
If I'm sure a page exists, but I simply can't remember the name exactly right (a common occurance with this page) I use UseMod's search facility, either with a term that I'm fairly certain is a substring of the name, or some other words that seem likely to be on the page. If I see the page-name in the search-results list, I copy it, return to the new page I'm making, and fix the link.
I other words, the one thing I don't do when making a link to a page is actually go and look at that page first!
After I've made the link, I do go to the page, of course. I go to remind myself of what I actually said there; and to see what other interesting connections it has. And probably to add a reciprocal link back to the new page I've just created. But that's later.
Now, it seems to me, unlike the blog case, there is no continuity between what I do when I make links in a wiki and what I need to do to use purple numbers. In order to find the purple number I'd need to be looking at the target page first, and then making the link to it afterwards. Now considering it's fairly typical that when I create a new 3 - 4 paragraph page, I'm embedding 2 or 3 links in it, plus another 3 or 4 on the "See Also" list at the end, and it becomes obvious why I'm fairly unlikely to make links to purple numbers within these pages. I don't want to visit all those pages before writing this one.
OK. But let's not make this a negative post. Given all that, what are the positive suggestions that could be made? Remember, my issue isn't with finer-granularity addressing or purpleness. It's with the "number" part of things ie. that paragraphs have arbitrary, purely syntactic markers attached to them.
But we could imagine markers that weren't meaningless. For example, if I could make a link which simply included the first X letters of a paragraph to be matched. If I could link to the previous paragraph in this post by making a link like this :
Now, there is a continuity with how I already make links with known pages. For very well known pages, it's not beyond the realms of possibility that I remember that a paragraph criticising the main proposition of a page starts with "The counter argument against this is". Or that I start evolving within-page conventions like "summary : ", "results : ".
(Aside : Actually I may try to implement this in SdiDesk if I can figure out how to make the HTML control jump to anchors within a page. :-)
Alternatively we can imagine an IntelliSense-type tool. For example, when I write the name of an existing page, a little drop-down, starts floating above my cursor to offer me a selection of paragraph beginings on that remote page.
I guess that's an example of the kind of tool Eugene is imagining.