This is a quick gander during a lunch break at work, but my (basic) impression has been that textons are the data, which carry with them some sort of material weight (digital or otherwise); scriptons are version of output; and the traversal function is akin to the operation of the query.

So, a bunch of rocks in a bucket - textons. Choosing them by size and arranging them in a weaving pattern - traversal function. Final "ideal" reading of the rocks laid out - a scripton. The reading is "ideal," of course, because narrative theorists are quite uncomfortable talking about the 'real' author or the 'real' reader - they get to be outside the brackets in those fun diagrams. So, yeah, I'd imagine a token nod to differences in readings and a way of removing that rather unwieldy topic from the equation.

I would guess that the algorithms and code that operate as the browser's html parser would be the "traversal function," with the html source as the texton and the document as viewed in the browser as the scripton. But, not a lot of variety there, unless you consider the differences in browsers, people's operating systems, monitor sizes - I guess you could keep going from there and include lots of stuff in the traversal function (or would it be in the materiality of either the texton or scripton?).

That accords with my basic sense of the terms as well. I guess what I'm getting at is that the dual nature of "texton" and "scripton" encourages one to think of only two layers/levels in the machine architecture. Of course we all know that the rabbit hole goes much deeper than that, all the way down to machine language and ultimtaely bit represenations on some physical storage media. What I'm wondering is how to apply Espen's terminology when the ground is always shifting (see my "update" paragraph). My sense is that his vocabularly *should* be able to account for the multiple, seemingly always just-out-of-reach spaces of the deep machine ("like city lights receeding") but I can't quite think my way there given the very spare nature of his presentation.

I think I see - are you concerned, then, that the model is in line with a kind of 'flickering signifier' mirage (and all of its non-material associations), where the "text" is theoretically comprised of only the surface level development of texton, through transversal function, to scripton? So, in a stab at a concrete example, if we're talking Uncle Buddy's Phantom Funhouse, to what layer of the texton to scripton process do we attribute that hypertext's incompatibility on machines other than Macs (and/or the MacOS software)?

And if we were to stick it in an emulator, well, how do we account - theoretically - for that? All the way down, through the various processes.

I guess to stab from another direction, my question would be (and this is meant in earnest, not as dismissive): what kinds of theoretical questions would we want to answer that would require a certain levels of depth in the, as put it so well, receding city lights? E.g. I know that in some pursuits, knowing what kind of ink Jane Austen's first edition of Sense and Sensibility is scribed in, or the watermark of her paper, is absolutely crucial for a theoretical point. In other arguments, not so much - the tradeoffs and limitations of our profession.

I can certainly see some points where this may prove quite valuable, especially when considering versions of programs that are under the same title, but are for different platforms. Currently, I'm playing Prince of Persia: Sands of Time on both the PS2 and the Gameboy Advance. The narrative is similar; the gameplay echoes across platforms. But clearly there are some real differences, just based on the limitations and abilities of each platform, and I'm not sure how the textons to scriptons accounts for that, or if it even does.

So, to ramble full circle, I think I'm back to repeating your question, and answering with an uncertain shrug.

"you concerned, then, that the model is in line with a kind of 'flickering signifier' mirage (and all of its non-material associations), where the "text" is theoretically comprised of only the surface level development of texton, through transversal function, to scripton?"

Or any level of traversal: assembly to machine language would do just as well as an example. But labelling one a texton and the other a scripton seems to unnaturally stabilize what is in fact an ongoing symbolic cascade--the same level (to stick with that image) of textuality can (and is) simultaneously *both* scripton and texton throughout the most mundane operations of the computer.

"what kinds of theoretical questions would we want to answer that would require a certain levels of depth in the, as put it so well, receding city lights?"

That's a good question, indeed *the* question (or my question, anyway ;-). One answer might be, given our current practices of engagement with electronic textuality, relatively few--and most of those in the specialized realm of preservation, migration, etc. Another answer might be, if such knowledge is deemed largely irrelevant--when it is so _basic_ to the operation of the machine itself--does that say more about our own critical limitations and the nature of the rather narrow questions we are bringing to the practice of electronic textuality? There was a time, of course, when knowing the kind of ink Jane Austen used was deemed all but irrelevant to literary stuedies, and of course, as you point out, in many quarters it still is. And that's okay, one can do interesting enough work anyway. But--to offer the counter-example--Viscomi and Essick argue for one *specific* technological procedure in Blake's method of color printing that has *direct* implications for our most basic critical conceits about Blake. To what extent does John McDaid's decision to work on a Mac contribute to the model of electronic writing on display in Uncle Buddy?

Matt, your agencement of the quotations helps one understand that

a Texton is not a Text
and
a Scripton is not a Script

Indeed as you have laid out the quotations, one associates

scriptons with strings as they appear to readers
and
textons with strings as they exist in the text

If a texton is the equivalent to a _state_ and a scripton to an _instruction_ there may be a machine-centred way of figuring the relation between them. A scripton could inscribe an instruction to simply note a state. Or if you wish, call an involutive transversal function. That is, such a scripton would present a script that indicates to the reader to do nothing with the current string. But how would a reader get to the next string or even navigate along a string?
It is now obvious that it is incorrect to characterize a scripton as calling a script or invoking a function. So we reread the sequence of quotations and note that a _text_ consists of three entities: scriptons, textons and transversal functions. Note that I in this restatement I have pluralized the transversal function. Aarseth's text seems to be pointing towards the existence of transversal functions that "hide" scriptons and generate textons from scriptons. There is a hint that scripton erasure could be productive. Indeed if the scripton is not consumed the texton cannot be made manifest.

This is all a deviation that veers into the political economy of cultural practices. All as a prelude to asking yet another set of questions: is a traversal function a scripton or a texton? how is a traversal function accessed?

This proposed viewpoint orignates in a reading of a biography of Turing. For some narratological implications of _configuration_ as either instruction or state, see
http://www.chass.utoronto.ca/~lachance/S4D.HTM#turing2

I’ll apologize now for the length of this posting. I’ve had trouble with these terms from the beginning, mostly thanks to Roland Barthes and his “From Work to Text.” But anyway, I’m going to try to get my head around the question and the hopefully present something useful. Our problem is, in a computer, *everything* is represented over and over and over again. So, walkthrough of an example of coding and recoding could be helpful, I hope.

Round 1: /This/ text is displayed in a charmingly familiar format to the reader when simply browsing. Underneath, it is iso-8859-1 encoded XHTML. This XHTML is transmitted as an HTTP 1.0 request, so even reading the source won’t give you everything, but for the moment we’ll abstract the HTTP details away. Anyway, we’ve got a set of Textons and Scriptons there: what I /see/ in Firefox are the sciptons, and the XHTML that Firefox renders (along with the HTTP overhead) are textons. There’s already lots of meat there, with the variety of XHTML used (1.0 Transitional) and the chosen character encoding, that it was sent over HTTP and not gopher or Telnet, as well as the decision my browser made on font, size, color, etc. But that wasn’t so bad.

Round 1.5: The XHTML really doesn’t tell the computer much about presentation. It’s really just a set of instructions to the browser about how to render the page. And the browser itself doesn’t really know all that much about the mechanics of the process either. So this time the browser converts the XHTML into system calls to the windowing system for things like rendering fonts, making scroll bars, and all that jazz. So maybe we have another round of scriptons and textons, but we have some ambiguity: what do we call the system calls that the XHTML is translated into? This is the traversal function for this level, I’d say. We’re not done though.

Round 2: XHTML is represented in my computer as transistor states in RAM. These transistor states are processed by my CPU care of those system calls in round 1.5 and then transformed into a new set of transistor states there, then sent across the front side bus as some sort of series of pulses to a graphics card, which then translates those pulses into again new pulses into my monitor, which translates the pulses into streams of electrons in my CRT and eventually are realized as light which I see and understand as the webpage. So we have a new set of textons (the transistor states) and a new set of scriptons (the photons that I think of as representing a page?). The traversal function this time is really weird and involves a lot of confusing BIOS and low level C stuff that I don’t understand well enough to talk about. However, there’s more to it than this. We’re already getting mucked up – what do we call the XHTML code this time? We’ve already set the textons as the transistor states, and the scriptons as the photons that I’m seeing. However, the XHTML is also both texton (in that it inspired the transistor states) and scriptons (in that it inspired the photons). Maybe we can call it a virtual-on, but that doesn’t mean anything. Let’s try to get more confused first.

Round Pi: Our friendly XHTML for this page lives on a server (www.otal.umd.edu). To get from this server to my computer, it needs to be converted to packets for TCP transmission, and then its put back together on my computer. However, now our reader is a computer, not a person, and it’s a weird circular process of XHTML --> Packet --> XHTML. Since textons and scriptons are defined in terms of the receiver (as far as I can tell), we’ll say the packets are textons and the XHTML is scriptons here. This is getting really pretty weird. Plus, we have another level lower than this of the Ethernet cards, routers, and all that fun reconstructing and recoding everything in the process. So, this whole terminology is getting really perplexing.

What do we do from here? First, we have to decide: are all the levels of coding (after level one) really counting in his framework? Since the transistor states end up as verbal information, are they in a sense textual? And can a CPU be a “given observer”? I’d say sure. Secondly, there’s the problem of levels. I think the problem is similar to that posed by the “Mood” section of Genette’s _Narrative Discourse_. In “Mood,” Genette defines several narrative levels: Diegesis, metadiegesis (or sometimes extradiegesis), and intradiegesis. However, in a relatively complex narrative, any section of the narrative could at any time be two or maybe even three of these terms, depending on what level of the narrative you define as primary. And so it is with texton and scripton. One needs to define a level of primary (or first) textuality and then work from there. One could then number the levels and use a subscript notation to make things clear. So our primary level could be the transistor states, and that would be level 0. Then, a level up would be XHTML and system calls. Then, the level above (or maybe an orthogonal level?) would be XHTML and the screen layout (the most superficial level). I’m not entirely sure how useful that is to anyone, but it at least starts to address the problem as I understand it.

Excellent, Matt. Of course you left out stylesheets, caching, swap space, virtual memory, the low-level disk geometry on the original server . . . ;-)

You're quite right; there's quite a bit to it. I think going through everything in detail might be instructive though. I've set up a page over at SoLaSI so that this can be collaborative.

http://www.solasi.org/moin.cgi/CodedAndRecoded

I've started the process with a few levels of detail and I'll probably add more later, but everyone please feel free to add and edit to the page.

Great idea for a wiki, Matt. I've posted a note in the comments over at GrandTextAuto. Will contribute myself soon.

I'm not quite so sure that the relation between texton and scripton is specular. They do not mirror each other. They are not reverse images of each other. Neither are they ontologically equivalent.
There are indeed particular meaning-effects that are engendered from the forms registration (recording, inscribing, physically storing) can take. [For more on the ontology of registration see Brian Cantwell Smith's On the Origin of Objects.] However, if one wants to avoid positing a perpetual regressive series or an eternal mise en abyme, one might want to stop and ponder if each act of registration is indeed equivalent.
The language of the passage from Aarseth is very precise. A text contains textons, scriptons and a traversal function. Note _contains_. Not _consists of_. A text in this sense is a bordering space.
A text is generated.
Scripton at t1, scripton at t2, scripton at t3 (faded, fade, fad, fa)
A text is regenerated.
A text is degenerative: a text is a way into another text. That way is always a distorting way.
The Textons of a Text are not accessible through the Scriptons of that Text. Textons are accessible via the scriptons read off another text. My desire to pluralize the traversal function returns. But it is now perhaps clearer that alternative traversal functions operate on copies. The act of reading creates a copy in memory. What counts as a copy? what counts as novel? The answer depends upon a reading of the temporal sequence, a parsing for the ruptures. A scripton at t4 may be taken as a different scripton. To take a scripton, whether emergent or continuous, is to contain it in a text. A text is a machine capable of modeling other machines. Barthes does begin S/Z with a key section entitled "Evaluation" whose first sentence reads in one English translation "There are said to be certain Buddhists whose ascetic practices enable them to see a whole landscape in a bean."
Copies put pressures on texts just as a genome shifts with bioforms instantiated from it. The bidirectionality of influence between copy and text ( a very non-Platonic view of text ) depends upon theorizing the inaccessibility of the textron.