![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
Entry tags:
Rhizomes, Databases, and Heidegger
The title makes this sound boring, but bear with me. It's going to be ridiculous and really boring instead. Also I am submitting this as homework but I am writing it primarily for this audience and my class can deal with it. (Also this is an attempted edited version from what I posted last night, particularly in the last two paragraphs.) So:
Deleuze and Guattari, in their introduction to A Thousand Plateaus: Capitalism and Schizophrenia, discuss three different ways of organizing knowledge (and thus of organizing the world): the arborescent model, the radicle system, and the rhizome. The arborescent model is like a tree, growing from a trunk into a branching structure. The radicle system works like the part of a seed from which all of the roots grow --- there is one point of articulation from which everything else is developed. The rhizome is, well, a rhizome --- while it sends out occasional roots or stems, all of the parts of the rhizome are just rhizome, and do what they do. (Ginger root is not actually a root and is a good example of a rhizome. If you cut parts of it off, the edge will turn into edge. It all just... gingers.)
In trying to understand this and what I might be able to do with it, I started thinking about data structures. The arborescent model is a common way to construct a data structure; you start from some point, and then fork in two directions, and each of those points can fork in two directions, and so on. You can even have points fork in more than two directions if you want, but the way you move through the data structure is to traverse the tree and get to individual nodes of data (which I think are even called leaves if they terminate, but it's been a long time since my intro CS class). This makes certain types of searches incredibly fast/efficient because you can make certain assumptions about the way the data is organized --- at my job, our secret sauce database is based on this kind of structure --- but relies on having the entirety of the structure to work. If you disrupt the tree, the relationships between the leaves, and thus most of the leaves themselves, are inaccessible.
The radicle system strikes me as more like a lookup table or dictionary. The radicle itself contains paths to all of the pieces of knowledge within it, so you can say radicle[key] and you will get the value of that key. If you take a segment of the radicle system away, one of two things can happen. If your segment includes the radicle, you at least know what information you are missing, because you will have connections to some values and not to others. If you lose the radicle in your cross-section, you just have arbitrary data points floating without context, which may or may not help you at all. (There isn't much reason to take a random cross-section of a database that I know of, but we'll get to why this matters in the Heidegger part.) The radicle system is also called the "fascicular root" and D&G connect it to fascism because of the way power is bundled in one place (not just because of the wordplay). I don't think a hash table is fascist; that's pretty ridiculous. But the concentration of power and knowledge may be.
The rhizome, on the other hand, contains all of what's needed to generate itself in all of its parts. So every piece of the ginger knows how to send up a stalk, or put down roots. It's... sort of like self-replicating code? You can chop up the ginger and each piece of ginger will still be able to do what the ginger does. Now, they may put down roots or send up stalks in different places, but the overall algorithm doesn't changed; nothing about the organizational structure except the individual manifestation of stalk/root is changed. I'm not sure how this would be useful in a digital data structure (anyone know?) but it's interesting to D&G because of its non-hierarchical structure and lack of necessary organization. (This ties into the Body without Organs.)
Heidegger, in "The Question Concerning Technology," says a lot of really busted and technophobic things (including the ludicrous claim that a windmill does not take energy out of the wind but a turbine does out of the water, because Heidegger is wiser than all gods and scientists) but provides something useful in the idea of the standing-reserve. Essentially he criticizes modernity for treating the world around us as resources to be exploited rather than entities engaging in active being in relation to each other. He follows up on this in "The Thing," claiming that modernity/science looks at things as objects and to truly understand them we have to look at their fourfold (earth/sky, human/divine) nature and their enmeshment in all other things, particularly but not only systems of gift-giving and ritual.
When I put these things together, I don't see rhizomes versus arborescent structures; I see entities that interact in rhizomatic, radicle-ish, and arborescent ways in their complicated enmeshment, but that sometimes have a dominant form of interaction. Rhizomes put down roots; trees sometimes end up with radicles; some trees can be cut and grow from cuttings. (Tree radicles: The mesquite trees in Tucson, for example, sometimes hit a point at the end of a branch with a sort of weird nodule that has dozens of branches coming out of it. I wish I had a picture, it's kinda freaky.) The body without organs --- the rhizomatic space, smoothness --- cannot be achieved, it can only be approximated. Even as rhizomes we have to put roots down to get nutrients, extend up flowers to reproduce, and those connect us to the world in non-rhizomatic ways. Organization is required for haecceity, or roughly (I still don't quite get this concept either), the distinguishment of one thing from everything else.
How I read D&G suggests that we need to not grow too attached to those connections, continuing to produce new and different connections rather than letting the stalks that provide for the flowers turn into trees or the roots turn into fascism. At the same time they're not suggesting the disconnect that upsets Heidegger so much --- in particular, their chapter on the war machine argues against the standing-reserve when it emphasizes the nomad's connections to and relationships with both the literal stuff of earth and to the State. The nomad figure's pack nature calls attention to the figure's multiplicity, but also to its internal replicability; if you separate part of a nomadic pack and put it somewhere else, it will keep doing what it does in a different context, just like lifting part of a river and putting it somewhere else will still produce a river, but one that looks different because of its relationship to different grooves and channels in the ground. In this way taking a cross-section of a nomadic pack is more rhizomatic than taking a cross-section of the State (or of a database).
I'm still trying to work out exactly how to make this useful. Thoughts?
Deleuze and Guattari, in their introduction to A Thousand Plateaus: Capitalism and Schizophrenia, discuss three different ways of organizing knowledge (and thus of organizing the world): the arborescent model, the radicle system, and the rhizome. The arborescent model is like a tree, growing from a trunk into a branching structure. The radicle system works like the part of a seed from which all of the roots grow --- there is one point of articulation from which everything else is developed. The rhizome is, well, a rhizome --- while it sends out occasional roots or stems, all of the parts of the rhizome are just rhizome, and do what they do. (Ginger root is not actually a root and is a good example of a rhizome. If you cut parts of it off, the edge will turn into edge. It all just... gingers.)
In trying to understand this and what I might be able to do with it, I started thinking about data structures. The arborescent model is a common way to construct a data structure; you start from some point, and then fork in two directions, and each of those points can fork in two directions, and so on. You can even have points fork in more than two directions if you want, but the way you move through the data structure is to traverse the tree and get to individual nodes of data (which I think are even called leaves if they terminate, but it's been a long time since my intro CS class). This makes certain types of searches incredibly fast/efficient because you can make certain assumptions about the way the data is organized --- at my job, our secret sauce database is based on this kind of structure --- but relies on having the entirety of the structure to work. If you disrupt the tree, the relationships between the leaves, and thus most of the leaves themselves, are inaccessible.
The radicle system strikes me as more like a lookup table or dictionary. The radicle itself contains paths to all of the pieces of knowledge within it, so you can say radicle[key] and you will get the value of that key. If you take a segment of the radicle system away, one of two things can happen. If your segment includes the radicle, you at least know what information you are missing, because you will have connections to some values and not to others. If you lose the radicle in your cross-section, you just have arbitrary data points floating without context, which may or may not help you at all. (There isn't much reason to take a random cross-section of a database that I know of, but we'll get to why this matters in the Heidegger part.) The radicle system is also called the "fascicular root" and D&G connect it to fascism because of the way power is bundled in one place (not just because of the wordplay). I don't think a hash table is fascist; that's pretty ridiculous. But the concentration of power and knowledge may be.
The rhizome, on the other hand, contains all of what's needed to generate itself in all of its parts. So every piece of the ginger knows how to send up a stalk, or put down roots. It's... sort of like self-replicating code? You can chop up the ginger and each piece of ginger will still be able to do what the ginger does. Now, they may put down roots or send up stalks in different places, but the overall algorithm doesn't changed; nothing about the organizational structure except the individual manifestation of stalk/root is changed. I'm not sure how this would be useful in a digital data structure (anyone know?) but it's interesting to D&G because of its non-hierarchical structure and lack of necessary organization. (This ties into the Body without Organs.)
Heidegger, in "The Question Concerning Technology," says a lot of really busted and technophobic things (including the ludicrous claim that a windmill does not take energy out of the wind but a turbine does out of the water, because Heidegger is wiser than all gods and scientists) but provides something useful in the idea of the standing-reserve. Essentially he criticizes modernity for treating the world around us as resources to be exploited rather than entities engaging in active being in relation to each other. He follows up on this in "The Thing," claiming that modernity/science looks at things as objects and to truly understand them we have to look at their fourfold (earth/sky, human/divine) nature and their enmeshment in all other things, particularly but not only systems of gift-giving and ritual.
When I put these things together, I don't see rhizomes versus arborescent structures; I see entities that interact in rhizomatic, radicle-ish, and arborescent ways in their complicated enmeshment, but that sometimes have a dominant form of interaction. Rhizomes put down roots; trees sometimes end up with radicles; some trees can be cut and grow from cuttings. (Tree radicles: The mesquite trees in Tucson, for example, sometimes hit a point at the end of a branch with a sort of weird nodule that has dozens of branches coming out of it. I wish I had a picture, it's kinda freaky.) The body without organs --- the rhizomatic space, smoothness --- cannot be achieved, it can only be approximated. Even as rhizomes we have to put roots down to get nutrients, extend up flowers to reproduce, and those connect us to the world in non-rhizomatic ways. Organization is required for haecceity, or roughly (I still don't quite get this concept either), the distinguishment of one thing from everything else.
How I read D&G suggests that we need to not grow too attached to those connections, continuing to produce new and different connections rather than letting the stalks that provide for the flowers turn into trees or the roots turn into fascism. At the same time they're not suggesting the disconnect that upsets Heidegger so much --- in particular, their chapter on the war machine argues against the standing-reserve when it emphasizes the nomad's connections to and relationships with both the literal stuff of earth and to the State. The nomad figure's pack nature calls attention to the figure's multiplicity, but also to its internal replicability; if you separate part of a nomadic pack and put it somewhere else, it will keep doing what it does in a different context, just like lifting part of a river and putting it somewhere else will still produce a river, but one that looks different because of its relationship to different grooves and channels in the ground. In this way taking a cross-section of a nomadic pack is more rhizomatic than taking a cross-section of the State (or of a database).
I'm still trying to work out exactly how to make this useful. Thoughts?
no subject
Anyway this makes me miss TST and thinking/talking about things with you and urrone.
Data structures and organs
First, I'm not sure how I feel about your comparison of information bases to data structures. I'm not sure how I feel about D&G's information organization models, frankly. At a minimum, it feels incomplete; there's no structure representing information that connects cyclically without assuming fractal reproducibility. There's seemingly no equivalent to the graph, unless you want to call it a subset of the tree, but those aren't really equivalent because trees imply hierarchy and graphs don't, or at least they don't in the same way.
Second, Heidegger is one of those philosophers that I can never quite take seriously. He seems to want to say a great deal without ever actually doing so, and his central tenets all seem to want to set up a theoretical divide between existence and execution, which is a fascinating idea but fundamentally unhelpful as a concept. I'm intimately familiar with the idea of the self-out-of-context, but I also know (as in speak from personal gnosis) its limitations. Context is inevitable, and even when steps are taken to consciously control it, it's also influenced by uncontrollable factors. This isn't to say that Heidegger should be discarded in every case, but even when he says something with which I agree, I question the steps he took to got there.
Third, I'm not sure your example of the nomadic pack is necessarily applicable, however illustrative it is. Context again being what it is, the nomadic pack has at least some radicle-style interaction with its environment. A pack suited for desert survival would probably be lost in the tundra, for instance. Also, subdividing the pack assumes that each member of the pack operates at a microcosm, but even in nomadic tribes we see some evidence of specialization that may prevent substructures of the pack from surviving isolated from each other. Arguably, yes, you could construct a cross-section that could survive separate from the whole, but you couldn't necessarily take any arbitrary cross-section and have the experiment work, which is what I think the rhizomal nature implies.
All of which really gets me down to the question you asked, which is how to make all of this useful, and that's where I have to say "I have no idea." I can only really go back to the scientists' questions at this point:
* What are you trying to model?
* Does the model you've made give you the ability to predict or understand phenomena in that on which your model is based?
no subject
A more radicle structure might be the kernel/userspace distinction in an operating system. The kernel is trusted, and ends up mediating almost all interactions between different parts of the system.
no subject
no subject
no subject
OK, I found my reeeeally old paper on Delany, in which I wrote the following:
In “Toto, We’re Back!,” a 1986 interview with the Cottonwood Review, Samuel Delany develops what he labels a “web” or rhizome, based upon the distance, or “mutual inadequations,” between language, desire, and “what happens,” the last being our behavior alone and with others (59).
and
“Aversion/Perversion/Diversion,” while it showcases Delany’s role as “storyteller” and the nature of “Gay Identity” explicitly, uses the rhizome of “Toto, We’re Back!” as an implicit code (119, 142).
Wow. That was... a different me. Reading one's early grad school papers is interesting. "Aversion/Perversion/Diversion" is in the essay collection Longer Views. You could also consider Times Square Red, Times Square Blue as a kind of analysis of space inspired by these concepts--I heard someone give a good talk on that, once. (Finally, but not all that relevant: "Atlantis Rose...", also in Longer Views, is the best piece of lit crit on Hart Crane ever.)
no subject
no subject
Noting that I don't actually understand what half your words mean, just on a style note, I think the analogy that roots turn into fascism came out of nowhere. Actually, the whole second half of that paragraph feels really rushed, especially compared to the loving descriptions of data models earlier in the piece.
no subject
no subject
no subject
At a lower level, perhaps an array of pointers? Cut off part and the rest still mean something, the same thing even if they point to isolated, autonomous functions, one of which could be "rebuild the array of pointers in this way".
I haven't read the books myself, but it sounds like they are trying to articulate the advantages of late-binding, message-passing systems over procedural design.
Heidegger sounds like he'd like Haskell; where doing is being and only possible in relationship to other world-entities, knowledge exists only in liminal spaces between those entities, and entities can change at any moment. In functional programming there are still plenty of embedded assumptions, but none of them correctly apply to what the system looks like.
D&G sound like they're Ruby folks: just because the last time I called X doesn't mean this time the result will be the same: maybe the entity I'm talking to has different data or is an entirely different entity that happens to also be able to do X.
It's a little weird to focus only on knowledge, in isolation from the ways that knowledge is used. Packaging data and computation is nearly always a win. One thing CS seems to point to is that the format of knowledge is only important in terms of speed. If it articulated, there is a way to find it given sufficient time: what is important for major leaps in processing power is the organization of agents, sharing/subdivision of information and abstraction such that complex tasks become comprehensible. Having successfully, flexibly self-organizing components can answer questions where procedural code would, at the very least, be unmaintainable, whether they are looking at flat data or organized data structures.
no subject
no subject
no subject
Cloud computing is “in” because, now that fast Internet connections are cheap (by First World standards), it is often more cost-effective to contract out the whole messy business of keeping a physical server running, keeping it connected to the Net, backing up data, etc., then to make someone on your own staff responsible for all that. There’s probably something Marx-like to say about this trend, but I’m not sure how it fits into the D&G schema.
no subject
Holograms work that way, with most of the data available in little parts of the film.
no subject
no subject
The other thing is basically everyone's desktop has its own version of The Knowledge, and it differs from other desktops and then you have to reconcile the various desktops but none of them contain All The Knowledge, they all just contain Knowledge. Presumably in the background they all talk to each other and update each other, and I do not understand resolving conflicts. But at a previous workplace, a number of nerds were like "But this distributed model is the coolest thing ever and solves assorted problems" and looked down on people who were like "But I like having a server!" This seems different from the Cloud in how the user thinks about it, although I could be wrong in that maybe the Cloud is all just a version of this, not that I think anyone really knows what the Cloud is except people who are involved in lawsuits about it.
It makes me think of single point of failure and multinode installations and such, and I don't know what to make of all of it.
no subject
There is in fact still a database used in git (called a "repository") with a trunk and branches, but everybody has a copy of it rather than interacting with a central server. You would then also have a "working copy" of the source code which you've taken from your repository, and you make your changes in there. Once you're satisfied, you commit your changes to your copy of the repository database, which stores a record appropriately.
Then comes the clever bit. You can tell your copy of the database to synchronise itself with somebody else's copy: the two approaches are called "pull" (you get all change records they've added to their database and merge them into your own) and "push" (you send your new change records to them for merging). It is still possible to have conflicting changes which must be dealt with by hand, but that's an unavoidable problem really.
When a new person enters the project, they "clone" someone else's repository, which gives them a copy of the ENTIRE database according to the copy-ee. That means all the history of changes, trunk and branches, as well as the current state of affairs. It may not match the entirety of what ALL people have in their personal copies, but that's what the above synchronisation is for!
In practice a lot of projects have a "master" repository hosted on a server somewhere which everybody synchronises with, but that is more of a convenience measure than anything else. Anybody's personal copy of the repository could become a "master" copy if it was nominated to be so: there is no technical distinction, only a social one.
I have been thinking about this sort of thing a lot lately, I don't think we make enough use of this approach in computing. It seems similar to the way we manage knowledge amongst humans?