An unfinished work

If you click around until it breaks, your browser becomes a text editor. The source code of this entire software is visible on every single page. These observable symptoms of failure reject the pattern language of institutional production around art to the extent that the modernist project for an art form is still an object of self-repair. There is no business model. There are no paywalls. Anyone, and I mean anyone, can make changes to this server, this site, this page and this paragraph.

In Early History Of Smalltalk, Alan Kay explains how after seeing Seymour Papert's work with Logo Language he grew to see home computing less as a "personal dynamic vehicle" and more as a "personal dynamic medium." He understood the relationship between the system and the environment as a simulation. The pattern language for personal computing the Sun Microsystems pitched to Microsoft in the 1980's or the 1990's interpreted Kay's Smalltalk language through the relationship between the model, the view and what the user could control as a system of symbols.

Luhmann wrote about art as a kind of incursion between perception and communication that reproduces the real and originates it at the same time. Simulation can do this by exploiting the architecture of self-reference, but only if autonomy is the solution to the complexity that new information introduces.

> In this sense, the autonomy of art is always operative Autonomy, therefore self-limitation of the work of art. At the same time, however, autonomy is also the object of the reflection of the art system [→Self-reference]. With Romanticism, reflection manifests itself as artistic criticism and gives the theory of art the status of self-description of the system. Criticism is not a program for creating works of art. It is a program for second order observation that formulates the autonomy of art as a constant negation of art itself, as a necessary and impossible transcending of imagination.

Aaron Swartz's unfinished work was the programmable web. I believe the internet, and by extension, the world, would be different were he alive to complete it. With this system of servers, *.by-sa.cc I have inverted the system of names otherwise administered by ICANN at the top, the server farm in the middle and wildcards at the lowest branch of privilege in order to put the work back into the hands of the artists who own it. Swartz's 2009 build plan did not inspire my implementation of the wiki federated since 2011 but this document anticipates the intelligibility and legibility of user-editable interfaces as defined by the world's fifth most visited website.

A federated wiki is an http browser and a license. Subdomains on *.by-sa.cc are issued on demand and administered by the Creative Commons. The software itself is under an MIT license. Pages are meant to be addressable and linkable as pages in the open web, and the whole “lineup” navigation is URL-driven too. This is in alignment with Swartz's treatment of a programmable web in which URL's make sense, don’t break, and provide direction with durability.

The Smallest Federated Wiki project used normal web retrieval and simple endpoints (like page JSON), from the beginning. Now it publishes machine-usable indexes like sitemap.json that agents can crawl. This is somewhat less disciplined than Swartz' idea of a web semantics by design, but user choice through import + export lets you share without reinventing the wheel in any sane format: audio, video, code, mp4, pdf, graphs.

The federated wiki is built around data portability, not content sharing. Right now it is difficult for artists, art critics, art scholars and art students to access documents of contemporary art, and especially conceptual art, because these records are trapped in some opaque database. If sometimes a good sculpture is drinking beer with your friends, how do you license that?

As museums, specifically the Duchamp Research Portal, are making some records available as JSON files but lock in DMCA limitations as a condition of the index, mirroring these resources and making them available freely is necessary to preserve the conditions for open scholarship and open data. Federated Wiki is free to run/modify/redistribute and easy to maintain: page JSON + sitemap.json + links provide a crawlable structure, so you can build querying tooling on top in advance of AI as both agent and interface of the emerging web.

Swartz' idea of a “query at scale” makes CC-BY-SA + simple JSON + CORS in a browser necessary. A database, even a distributed database, consists of not much more than pulling one record from one API route. Asking a question that forces the system to merge, pull and request changes from the user end is slightly more elegant. Swartz said APIs only answer the questions you already know how to ask, but queries let you ask broader stuff.

This site models, mirrors and pushes the medium of the art record. When I'm not hosting artifacts (text, image, music) off of the same cloud server that issues subdomains, I'm pointing at an existing database that can and must be served by federation. This site, done right, will eventually stretch across multiple endpoints—one site is hosting the work of Simon Denny, another, an archive of 1960's art finance experiments. Only a human art historian would connect them, but being able to predict the next token is not how art history gets made.

Making cross-endpoint queries fast (so they don’t take forever) is hard and still an open problem. Joining datasets to query at scale involves the complexity of filtering art from content, before or after certain years, when most data online was published in the last 25 years. Multiple endpoints include not just online archives, legacy archives vis a vis experimental, counter-institutional resources, artist's websites, word of mouth communities and defunct BBS boards, and of course, brick and mortar commercial galleries.

On Federated Wiki reciprocal license is the identity system. Wildcards are the ultimate identity system. When malicious actors appropriate these services they do so in all four dimensions of ownership, authorship, site-specificity and prior availability.

The operations of the art system are observations oriented to works of art.