dir File Systems: The Original Hypermedia .as Essay .property description text A vision for a universal, accessible hypermedia system inspired by the file system paradigm .property cover image Trees .property source id cover.webp info Created time 2025-01-17 info Status text Draft anchor Cursed by aesthetics text HTML is a hypertext language created in the 90's by Tim Berners-Lee to power documents on the web. Its design was derived from SGML, a language extensively used in print production. Moreover, most document systems of the time were heavily oriented towards print. quote Idealliance on SGML .property source text SGML revolutionized print production workflows by combining typesetting with data processing for publishing sectors such as military, aerospace, automotive and law. .property by name idealliance.org text I argue that the influence of printed media on the design of HTML was essentially what made authoring web hypermedia evolve to require consideration of structural layout and formatting and, as a result, HTML defined a medium with no inherent character. text At first glance, HTML 1.0 tags seem to consist solely of semantic elements, and CSS wasn't introduced until years later. However, a closer look reveals subtle influences from print media concepts that have crept into its design. link HTML 1.0 Tags .property source id https://cetus.sakura.ne.jp/htmllint/tagslist.cgi?HTMLVersion=html10 text A strong example is the distinction between heading levels (`

` to `

`), a typesetting artifice used to emulate hierarchical structure through variations in text size. Another example is the presence of `` and `` tags, again features traditional to printed documents. text Such seemingly innocent and inconsequential design choices, in my view, were the breach that opened HTML to a series of revisions introducing even more layout and typesetting features. Eventually leading to its coupling with a full-fledged layout language. text As a consequence, authoring HTML documents became an artisanal craft, demanding meticulous decisions about positioning and decoration. Just look at the rich tapestry that was the 2000s web. Although this elevated HTML to highly versatile and powerful technology for creatives and specialists, it also introduced barriers to accessibility, making it less approachable for the average user and creating a layer of complexity that limited broader participation in the web. text When we think about other types of media, such as text, photography, or film, it’s easy to infer the character of their artifacts. If someone says they’re going to show us a picture, we expect to see a rectangular artifact depicting a place or object. If it’s a physical picture, we expect it to be made of pigment on paper. If it’s a virtual picture, we expect it to be made of pixels. text The depicted place or object does not interfere with the substantive, constructive character of the medium. All digital pictures are rectangular arrays of pixels. The same rationale extends to film, text and audio. A proper medium have innate and stable constructive and reproductive character. text The same cannot be said for hypermedia as it exists on the web today. With the cartesian, stylistic approach having brutally usurped the semantic, data-oriented foundations of authoring, it is no longer possible to reliably infer how a given hypermedia artifact will be structured, let alone how it will function. In my view, this undermines the status of web hypermedia as a proper medium. text If you were to tell me that you could, in fact, infer the qualities of unseen web hypermedia artifacts ahead of time, I would argue that this is the result of learned conventions and agreements people had to establish to make sense of the medium. But those agreements are superficial and uncertain. text What’s curious is that hypermedia, as encoded by HTML, actually does have a stable and inherent character, but for computers. It has a definite specification and regular structure. A computer can always infer the properties of HTML documents because it is a language made first and foremost for computers. text The consequence of Tim Berners-Lee not properly factoring in sociological, semiological, epistemological, and ontological considerations into his hypermedia format is that every artifact produced with it ends up reinventing the medium itself. It's cursed by aesthetic – by letting itself be steered by aesthetic, it has no aesthetic at all. text This is the fundamental reason we need entire industries focused on producing and maintaining web hypermedia. Independently authoring and distributing web content is a phenomenally complex endeavor. It’s what gave rise to content platforms, which are essentially hypermedia service providers capitalizing on the web’s complexity and unpredictability by offering a stable, understandable multimedia environment that humans can engage with. text The conclusion we can draw from this analysis is that the influence of print media on Tim Berners-Lee's invention and its subsequent improvements has led to the definition of a category and substrate, but not a fully realized medium. Just as the tangible instances of print media are embodied by books, magazines, newspapers, and so on, instances of hypermedia are realized through what we call content platforms, with each brand representing a distinct instance of the medium. text Ultimately, that's why the web is constantly being renovated, yet also constantly dying. It's our failure to conceptualize, formalize, and promote a humane, generalized means that would allow people to freely and openly participate on the web, unencumbered by proxies and intermediaries. anchor Cursed by complexity text An even worse curse that affects not only independently created HTML media but also platform media is the complexity inherent to dynamism. text When you define a medium whose materialization depends on continually sourcing external systems, you inevitably bind information to transient paradigms, making the media fragile and compromising its longevity. The knowledge contained within its artifacts becomes invalidated as soon as one or more of those interdependent systems go out of service or out of fashion. text Moreover, dynamic content heavily contributes to the deterioration of the already fragile model of digital information, further inducing the preconditions for what's been called the Digital Dark Age: the loss of knowledge triggered by technological obsolescence. link Digital dark age .property source id https://en.wikipedia.org/wiki/Digital_dark_age text A catalyst to the ever-growing dynamism in hypermedia can be attributed to the progressive convolution of its supporting environments, which no longer prioritize knowledge encoding and distribution as a first concern, instead becoming increasingly oriented towards the role of application engines, thus better suited to transactional, business needs. text To borrow vocabulary from economics, the difficulty to trade hypermedia across systems renders it unsuitable as a reliable commodity. This is striking because, while we consume it as if it were a commodity, the entangled mess of the underlying representations makes archival and redistribution in native form costly and ineffective, leading us to resort to taking pictures and screenshots of hypermedia as a means of preservation. anchor Hierarchy is dead. Long live hierarchy! text If you read enough papers on the design of information systems you'll perceive a trend of dismissing hierarchical systems as a competent solution for knowledge organization. Notably, Tim Berners-Lee did exactly that in his 1989 paper introducing the web. quote Tim dismisses hierarchical systems .property source text This is why a "web" of notes with links (like references) between them is far more useful than a fixed hierarchical system. .property by name Information Management: A Proposal. Tim Berners-Lee, 1989 file Information Management: A Proposal .property source id bernerslee1989.pdf text The argument goes that you cannot model and capture the complexity of the real world within hierarchies. quote Tim establishes the need for links .property source text | Many systems are organised hierarchically. The CERNDOC documentation system is an example, as is the Unix file system, and the VMS/HELP system. A tree has the practical advantage of giving every node a unique name. However, it does not allow the system to model the real world. For example, in a hierarchical HELP system such as VMS/HELP, one often gets to a leaf on a tree such as | | HELP COMPILER SOURCE_FORMAT PRAGMAS DEFAULTS | | only to find a reference to another leaf: "Please see | | HELP COMPILER COMMAND OPTIONS DEFAULTS PRAGMAS" | | and it is necessary to leave the system and re-enter it. What was needed was a link from one node to another, because in this case the information was not naturally organised into a tree. .property by name Information Management: A Proposal. Tim Berners-Lee, 1989 text The passage makes it clear that the central problem prompting Tim to conceptualize the web was the need for an overlay system enabling fast jumps between hierarchical nodes, essentially, an escape hatch to bypass traversing trees in a strictly ordered, top-to-bottom manner. text So it seems that, strictly speaking, it's not the case that hierarchies cannot model the real world, but hierarchies alone are insufficient to model complex topologies. Still, hierarchies are undeniably doing the heavy lifting at the base layer of systems; this is true both for the system exemplified in Tim's passage and, ultimately, on the web at large. text If we wanted to be even more strict, we could say that, in the context of the quoted passage, it was not the case that "information was not naturally organised into a tree", but instead that "information was not naturally consumed as a tree". text There’s no escaping hierarchies. To appreciate this as fact, we need only consider that order is the essence of hierarchies. The mere existence of firsts and lasts, befores and afters, constitutes basic rankings and, therefore, hierarchies. What makes orders intrinsically unescapable is the fact that we exist in a reality with a temporal dimension and are wired to perceive time linearly and progressively. text Time, and by extension order, is also what founds language and communication as serial systems of symbols and signs. So even when confronted with non-linear or hyperdimensional systems, we'll inevitably project or derive order from them, given that the seeking of order is our ultimate, inescapable bias. image The parametric and spatiotemporal nature of language .property source id notation.webp text The non-linear nature of hyperlinks, then, can only be fully appreciated in a conceptual sense. Outside of the mathematical, we perceive hyperlinks no differently than the flipping of a page – a lateral movement along the same axis – which is exactly how browsers encode navigation, with previous and next buttons. text Beyond that, the entire infrastructure powering not only the web but also the internet is largely composed of ranked subsystems. For instance: the TCP protocol with ordered packets, the IP, DNS, and URL systems with hierarchical namespaces, file servers built on top of hierarchical systems, and HTML and the DOM, which are based on tree structures. Additionally, most websites follow a hierarchical content structure, with ranked pages and subpages. text It's not too far-fetched to say, then, that the web is a collection of hierarchical subsystems enabling the interconnection of hierarchical systems. anchor One and the same text Let's rewind to 1965, before the invention of the Web, when Theodor Nelson introduced the concept and definition of hypertext and, more broadly, hypermedia. quote Ted's original definition of hypertext .property source text Let me introduce the word "hypertext" to mean a body of written or pictorial material interconnected in such a complex way that it could not conveniently be presented or represented on paper. .property by name A File Structure for The Complex, The Changing and the Indeterminate. Theodor Nelson, 1965 file A File Structure for The Complex, The Changing and the Indeterminate .property source id nelson1965.pdf text This characterization implies that hypermedia must have at least a dimension of depth, otherwise it can be losslessly represented on paper. Another way to look at it is by interpreting hypermedia as non-linear media, which simply means there's no single path that can traverse the whole structure. text So, for a medium to be a hyper, it must be multidimensional and multimedia. text Guess what? Hierarchical file systems fully comply with this definition: directories provide the multidimensional, non-linear qualities, while files confer the multimedia capabilities. text Even if we were to demand that hypermedia must also possess inter-linking features, file systems would still comply, at least the modern ones, which offer the ability to reference a file across multiple locations, not unlike hyperlinks on the web. text This begs the question: if we always had hypermedia with directories and files, why hasn't the web evolved into a mesh of interconnected file systems? Why isn't a website just a remote directory on someone's computer that we can explore via a file browser? text Well, it is sort of like that, but with extra steps. anchor A semiotic gap text Perhaps it was Tim's prejudice against hierarchies that blinded him to the opportunity of exploring the file system model as a more streamlined and robust way to conceptualize and implement the web as a medium. text Perhaps, as a scholar and scientist, it was his bias toward print media leading him to search for a "better version of paper", and file systems look nothing like articles or books. text Or, much more likely, it was that file systems are inseparable from operating systems, so it would be unrealistic to expect people to update their OS just to use this new thing called "web". text Nonetheless, the file system is such a powerful paradigm that it allows us to implement an overlay file system, following whatever model we want, directly on top of files. This is how database systems are implemented. text But regardless of the technical details, what I want to highlight is how closely the model of files and directories resembles a stereotypical hypermedia model, essentially requiring a minor semiotic gap to be bridged in order to blend one into the other. The gap I'm referring to consists of two missing qualities in the high-level interface of file systems. text The quality of inlining. .level 1 text Directories merely list the names and icons of entries; they lack the ability to render each entry's contained media directly in the explorer application. Instead, a file system requires a specialized application to visualize each type of media. text The quality of ordering. .level 1 text Rendering files inline wouldn't suffice to make directories readable like documents. For that we would need a way to tell which file comes first, second, etc. text If you enhance a traditional file system with these qualities, it would gain the emergent ability to emulate common content stereotypes found on the web. text You could emulate a blog post with a series of inline texts, or emulate a photo feed with a series of inline images. More generally, by the intentional arrangement of directories and inline media, you could emulate whole websites. text But it would be fundamentally different from hypermedia on the web. The hypermedia rendition that would blossom out of this file system would be characterized by its universality – its innate quality of free-form multimedia composition. And considering that this enhanced file system retained the handling of presentation as a systemic, native concern, the resulting hypermedia system would free users from stylistic concerns and fashion-driven pressures, making it a pragmatic medium with a purely informational focus. text Users of this system would still be able to express individuality through the media contents of their directories. Since creative expression wouldn’t affect the system’s primary usability layers, the interface would remain stable and consistent, evolving independently to address ergonomic and accessibility factors in a standardized, content-agnostic manner – as has always been the case with file system user interfaces, except with a broader scope. text It would matter little whether a given directory represents an "article" or a "site". These distinctions would gradually become irrelevant, as the character of the media would emerge from the structured nature of its contents, rather than from a superimposed aesthetic. Free-formed serial multimedia arrangements would itself be the dominant aesthetic. text And once you figured out the details to make this enhanced file system networked, you would have essentially created your own version of the web, and arguably a better one, with strong mediatic qualities. A hypermedia system so simple and standardized that it would have a much lower barrier to entry, because you wouldn’t need a second system to create content. You would create and manage content directly from the file explorer application, in the most natural way possible. text This version of the web wouldn’t require users to learn advanced computer skills in order to participate. Unlike the web we have, which is, for most people, effectively read-only, this version would create an inclusive, writable web where users naturally retain ownership of their data, thereby granting them digital sovereignty.