Purple Identification
May 11, 2005
In a fine bit of conversation, Phil Jones has responded to my response to Adina Levin responding to my response to Jason Kottke talking about the fundamental units of content on the web. (PLT)
Phil asserts that I'm hostile to the semantics of labels being bound by everyday behavior. This is not at all the case. Phil uses WikiWords as a good example. My hostility to free linking should indicate the depth of my love for WikiWords. I also love tags, but as Bill Seitz points out, tags don't point to a single resource, they point to a collection. (PLU)
Tags and WikiWords don't serve the same purpose as purple numbers. (PLV)
Tags and WikiWords are names people can use to label something or some things. They grant a certain power to authors: "I am calling this thing I am writing or this thing I am pointing to FOO". There is an expectation or hope of collision. (PLW)
Purple numbers are identifiers that grant power to readers, reviewers, annotators and commentators to indicate a specific piece of any content anywhere and make reference to or, more importantly, reuse it. If pieces of content had meaningful labels, imagine the difficulty of adding labels to every piece of content? (PLX)
For perspective: Tim Bray took a step down the slippery slope away from the shimmering shiny idealism of allowing the reader access to everything by implementing his (reference only, no support for transclusion) purple numbers in way that grants the author control over what chunks get the numbers. That's like letting a politician say "off the record..." (PLY)
By the way, a closer look at my gripe with URIs (PLZ)
Current implementations of purple numbers expose the identifiers (in both the numbers on the screen and numbers at all senses) but this does not need to be the case. Because they are unique (for now in a given suite of tools, but long term globally), persistent and stable they can have labels associated with them that resolve to the stable identifiers (Purple:DistributedPurpleNumbers for some references). The labels could be names like "mom's address" or sequences like those found in legal documents. All this is very much like the concept of a URI except that URIs, because they contain information about what they identify, fail to be persistent or stable and thus are not identifiers at all but labels posing (miserably) as identifiers. T (PM0)
shows that I'm okay with labels (which is what WikiWords and tags are) in general. I agree that presenting purple numbers onscreen as the numbers, and requiring users to manipulate the numbers is problematic. (PM1)
However given the goals (PM2)
- a high degree of flexibility in referencing content from anywhere, even content that will change its location on the network (PM3)
- referential control in the hands of the reader (PM4)
- versioned nodes (PM5)
there need to be persistent, unique, stable identifiers underneath whatever helpful interfaces will eventually exist between the user and the guts of the system. In simple wiki systems, the page name which people use to gain access to the resource which is the manifestation of that page on that particular system resolves to a unique identifier (inode) on the (extremely local) filesystem. In modern operating systems, you don't personally use inodes to get at stuff, but they are a crucial piece of the pie. Purple numbers, someday, can be thought of as inodes for individual chunks of content that transcend filesystems and local networks and can move around. (PM6)
Someday, perhaps, there will be nice libraries for them. In the meantime we have to come up with what those libraries will need to do by messing with some ick. (PM7)
Comments
(Just catching up on this thread.) (PMA)
As a general principle, I think any block should be able to be made referencable such that it can be linked to. Personally, I think it's very interesting to consider what might happen if you were to allow different people to simultaneously define what the boundaries are for the blocks. (PMB)
That said, an author or a primary editor is in a good starting position to be able to imagine what are the, generally, smallest, useful sized blocks in any document. And, doing so then allows others who wish to arrange things their own way to mostly just stack blocks, which is pretty easy in a lot of ways. (PMC)
Nevertheless, there will be some cases where someone wishes to extract part(s) of a block, and that's not as easy, but do-able. (PMD)
People used to build systems that did stuff like this, back in the pre-web SGML days. And, these were probably a big part of the direct precedent for fragments on the web. (PME)
Purple Numbers establishes a practice in the context of web pages, but older systems looked at this feature as something that had to have functions at all of the computer's reader, author, and manager levels. (PMF)
In other words, in today's terms, for example, we should look at how a web browser could be different if it could recognize purple numbers, or how an editor (e.g., blog or wiki editing screen) could be different... (PMG)
"Purple numbers, someday, can be thought of as inodes for individual chunks of content that transcend filesystems and local networks and can move around." (Q58)
You know, I've been busting my brain on this since '88. Working on a MIL-SPEC project I had to create a "family tree" that included every module in a (large and complex) aircraft landing system. The bigger part of my job was (apart from writing 3 system manuals) to track changes in system design documents ... including software, that flock totalled nearly 100 docs. (Q59)
What I imagined but never achieved was a system that tracked something like factoids. I think my "Participatory Deliberation" design has found a crack in the cosmic egg, but only by attacking from another angle. (The diff between a street fighter and a pro: the pro waits for the bell!) (Q5A)
I think semantic web's biggest problem is that it's dealing with absolutes. Because with ParDelib? I'm dealing with something smaller and fuzzier, something that's fault tolerant, I don't need 100% disambiguation. Collision? *shrug* Ok, so sue me. It's like going to the washroom that's 7 doors down and to the left instead of the one that's 2 doors down and to the right: it only matters if there's a dope-deal doing down. (Q5B)
cheers (Q5C)