At the bottom of somebody's personal site you find a small box: an arrow pointing left, an arrow pointing right, a name in the middle, and sometimes a die or a shuffle icon. Click the right arrow and you land on a different site by a different person on roughly the same subject. Keep clicking and you eventually come back to where you started.
For the broader context behind this business guide, read What We Lost When The Blogroll Died.
That is a webring. The format is old, it never entirely stopped, and people are building new ones. This explains how it works well enough that you could implement one, why the original wave collapsed, and how it compares with a checked register, including where the ring is better.
What this is, and who it is for
For someone who has seen the widget and wants to know what it is, and for someone considering starting a ring who would like to know which part is the actual work.
The wider history of discovery mechanisms is in the wider history of discovery. This piece is only about the ring.
The mechanism: next, previous, random
Strip out the graphics and a ring is a circular linked list of URLs plus three operations.
The data. An ordered list of member sites. Each member has a position, an address, and usually a name. That is the entire state of a ring.
The three controls, on every member's page.
- Next takes you to the member after this one in the order, wrapping around from the last to the first.
- Previous takes you to the member before it, wrapping the other way.
- Random takes you to any member, chosen without reference to your position.
How the controls know where you are. In the classic implementation the widget on each member's page is a link to a central script with that member's identifier attached, roughly of the form "next, from member seventeen". The script looks up seventeen, finds eighteen, and redirects you. The member's own page contains nothing but a link; all the logic sits in one place.
The modern alternatives. You can do the same thing without a central script. A small file listing the members can be published at a known address and fetched by a script on each member's page, so the traversal is computed in the browser. Or the ring can be entirely static: a published list of members, and each member's widget hard-codes the specific neighbours either side of them. The static version has no infrastructure at all and requires an edit on two sites whenever the membership changes.
What you would need to build one. A list of members in an order, a way to render three links on each member's site, and a rule for what happens when a member's site does not respond. That last one is not an implementation detail; it is the whole maintenance problem, and it arrives on day one.
Who ran a ring, and what the ringmaster actually did
Every ring had a coordinator, generally called the ringmaster, and the job was more than administering a list.
Carol Casey's article "Web rings: An alternative to search engines", published in College & Research Libraries News in 1998 (volume 59, number 10) and retrieved on 6 September 2026, describes the role from inside the period. "Each Web ring has a homepage that is created and maintained by the Web ring owner." "The ringmaster, as the owner of a Web ring is called, must have a good understanding of HTML and experience in creating Web pages." The design of the graphic and text elements that represent the ring on the individual member sites "is also the responsibility of the ringmaster". Casey also describes the recruiting — time spent "to track down potential sites, not to mention contacting Web site owners and, in some cases, convincing them to join the ring" — and what arrives once a ring becomes known: "a policy that controls the content and quality of the ring must be developed and more time spent evaluating these new Web sites".
Current rings publish comparable obligations in their own words. The XXIIVV webring's repository documentation, retrieved on 6 September 2026, states that the ring "is managed by" one named person, "but any member of the network is also welcome to join this repository as a collaborator to help manage new links and Pull Requests", and that "Websites without activity for over two years might also be periodically removed."
Casey's account covers the homepage, the recruiting and the content policy. The rest of the list below follows from the mechanism described above rather than from that source.
The work, roughly in order of how much of it there was:
Deciding who is in. A ring has a theme. Applications arrive that are adjacent to the theme, or on-theme but thin, or on-theme and commercial in a way the existing members would object to. Somebody has to decide, and every decision is about a specific person who will read it as a judgement of their work.
Keeping the chain unbroken. If a member's site disappears, moves, or removes the widget, the traversal breaks at that point. In a circular structure a break is not a local problem: it can strand everyone downstream of it. So the coordinator has to notice breaks, chase the member, and eventually remove them, which is a recurring obligation proportional to the number of members.
Moderating. A member's site changes over time. A ring is a public association between its members, and if one member's site becomes something the others would not want to be linked to, the coordinator has to act, and every option available is socially expensive.
Being the single point of failure. The list, the script, the domain it runs on, and the institutional memory usually lived with one person, unpaid, with no successor named.
Why rings collapsed
Five causes, overlapping, and none of them is about the format being outmoded.
The chain is fragile by construction. A linear list where each element points to the next has no redundancy. One dead member degrades the experience for everyone; several make traversal pointless. Link rot is continuous and silent, so a ring decays whether or not anyone is looking. The Pew Research Center's report When Online Content Disappears, published on 17 May 2024 and retrieved on 6 September 2026, drew "a random sample of just under 1 million webpages from the archives of Common Crawl", approximately 90,000 pages for each year from 2013 through 2023. It reports that "A quarter of all webpages that existed at one point between 2013 and 2023 are no longer accessible", and that "Some 38% of webpages that existed in 2013 are not available today". That measures pages in a crawl sample, not ring members, and it says nothing about webrings specifically. A ring is a list of pages, though, and it loses them on the same terms.
The maintenance was unpaid and recurring. Same shape as every other curated structure: intake is a one-off, repair is forever, and nothing funded the second one. That argument is set out in the cost of intake.
Central infrastructure was a single point of failure. Where traversal ran through one script on one domain, the entire ring stopped working when that domain lapsed or the service hosting it went away — even though every member site was still perfectly fine. A distributed structure with a centralised control mechanism inherits the weaknesses of both.
Personal publishing moved onto platforms. A ring requires each member to place a small piece of markup on their own pages, in a position that persists as the site changes. On a hosted platform the author does not decide what markup is available; the platform does. WordPress.com's own support documentation on adding code, retrieved on 6 September 2026, lists the HTML tags "All WordPress.com sites can use" — anchors, images and the ordinary formatting tags — and then states that "sites with a plugin-enabled plan and activated hosting features can use the following codes: embed, frame, iframe, form, input, object, textarea, script, style, link". It adds that "JavaScript codes typically begin with a <script> tag and may be used on sites with a plugin-enabled plan". That is one named platform on one date rather than a general law about the whole web. It is the shape of the problem all the same: a script-driven widget is not available to an author who has not paid for code, and any widget at all needs somewhere site-wide to sit, which is something the platform grants rather than the author arranges.
Traffic patterns changed. Rings assume browsing readers who arrived at one member and are willing to walk. Readers arriving from a query or a feed came for one page and leave from it, and a navigation widget at the bottom of that page is not part of their session.
What the current versions do differently
Three rings were loading when we checked them on 6 September 2026: the XXIIVV webring at webring.xxiivv.com, an IndieWeb webring at xn--sr8hvo.ws, and the Hotline Webring at hotlinewebring.club. They do not make the same choices as each other, and one of the three is closer to the classic implementation than to anything new. Four differences are visible across them.
Static or client-side traversal rather than a central script. The XXIIVV webring's instructions say that instead of linking to the directory, a member "can also link to the next link in the ring by adding parts of your site or domain in the hash of the request url", so the traversal is worked out from a list the reader's browser has already loaded. That removes the single point of failure: if the coordinator disappears, the ring keeps working with whatever the last published list said. Not every current ring does this. The Hotline Webring's joining instructions tell each member to link to /next and /previous on hotlinewebring.club, which is the classic central arrangement carrying the classic dependency.
Membership as a public file. The XXIIVV webring is joined by adding your entry to the index.html file in its public repository and opening a pull request. The ring's state is therefore a versioned file: inspectable, with a visible history, and forkable if the coordinator stops. Its documentation also names a succession path, saying the ring is managed by one person but that "any member of the network is also welcome to join this repository as a collaborator to help manage new links and Pull Requests". That converts a private list into a public artefact.
Automated liveness checking of members. The IndieWeb webring's front page states that it "automatically keeps track of (in)active member sites" and that "your previous and next neighbors are active in the webring". That addresses the chain-fragility problem directly, and it is the thing the classic implementation could not do without somebody watching. The XXIIVV ring handles the same problem by hand and on a slow clock: "Websites without activity for over two years might also be periodically removed."
A stated theme, and sometimes a stated ceiling. The XXIIVV webring publishes both. Its aim is "to share rich hand-crafted websites such as diaries, wikis & portfolios"; its criteria require at least ten content pages, an about page and your own domain, and exclude single-page sites and business homepages; and it states that "The webring will never exceed 256 entries, new entries will take place of websites that go dark." We counted 250 member entries on its page on 6 September 2026, against that published ceiling of 256. A ceiling is not universal: the Hotline Webring adds a site the first time somebody follows one of its links, and its published member table carried 860 rows when we counted it the same day. Where a ceiling exists, it keeps the repair burden inside one person's capacity, which is the same reason every other curated thing in this space is small — see why the current revivals stay small.
Ring versus register: what each is good at
Two structures, different properties. This is a comparison of mechanisms and not a ranking.
| Webring | Checked register | |
|---|---|---|
| Who consents | Every member, explicitly, by placing the widget | Only the publisher decides; the listed party need not be involved |
| What membership means | These sites agreed to be associated with each other | A named person examined this and it met a published bar |
| Who maintains it | The coordinator, plus every member maintaining their own widget | The publisher alone |
| How it degrades | Traversal breaks at the dead member and strands the chain | Entries go stale silently unless re-checked |
| Accuracy obligation | Implicit; the members notice when it breaks | Explicit; dated, and published as a claim |
| Social contract | Mutual, ongoing, and community-shaped | One-directional, with a removal path |
| Natural size | Small — bounded by repair capacity | Small — bounded by checking capacity |
The row that matters most is the first. A ring is an association its members opted into. A register is a set of observations made about parties who did not necessarily ask to be observed. Those are different kinds of thing wearing similar clothing.
What a ring can do that we cannot
Three things, stated plainly.
Consent. Every member of a ring chose to be there and can leave by deleting some markup. Our entries describe businesses that may never have contacted us. We honour removal requests free and quickly, at delist, and that is a weaker arrangement than never having listed someone without asking.
Community. A ring creates a mutual relationship between its members: they know who else is in it, they can see the association, and it is a group that a person can belong to. A register creates no relationship of any kind. Being in it is something that happened to you.
Distributed survival. A well-built modern ring survives its coordinator, because the list is public and the traversal is not centralised. If we stop, the register stops. The dates on the entries will make that visible, which is the only mitigation we have, and it is not the same as continuing to work.
What a register does that a ring does not is check. Ring membership records an agreement made at a moment; it carries no dated claim that the site is still what it was. Our entries carry a check date because the specific job we chose is keeping a short list true. Neither structure is the better one in general. They answer different questions.
If you want to join or start one
To join: find a ring on a subject you actually publish about, and look for its index page — the widget on any member's site links to it, and the index usually carries the theme, the membership rules and a way to apply. Read those rules before applying, because the routes in are not alike. Of the three rings we loaded on 6 September 2026, the XXIIVV webring takes a pull request against its repository and publishes a bar to clear first. The IndieWeb webring has you sign in with your own domain, and states that it "works with sites that support IndieAuth or RelMeAuth". The Hotline Webring has you pick a slug and add /next and /previous links, after which, in its own words, "When someone clicks on one of your links, you'll be automatically added to the webring." Even there somebody looks: the same page says "We look at every site that joins and ensure visitors land on the page with the webring links", and reserves the right to remove a site the webmasters judge outside their own bounds. Expect the coordinator to look at your site. That is the point of the ring.
To start one: the technical part is a weekend. The list is a file, and a static implementation needs no server at all. Everything after that is the work:
- Write the theme down before you accept anybody, tightly enough that you can decline something adjacent without it being personal.
- Decide the membership bar and publish it.
- Publish the member list somewhere versioned, so the ring survives you.
- Decide the check cadence and what happens when a member goes dark: how long a grace period, who chases, when they come out.
- Say what happens if you stop. Naming a successor, or publishing enough that somebody can fork it, is the difference between a ring that ends and a ring that quietly rots.
Nothing in that list is about code. All of it is about the recurring obligation, which is the part that decides whether the ring exists in three years. Our own version of that discipline is at our source ledger and across the independent web.
In Plain English
A webring is a circular list of sites on a theme. Each member's page carries next, previous and random controls, so a reader can walk from one member to the next and eventually come back round.
The technical part is trivial: an ordered list of addresses and three operations. The real work is the coordinator's: deciding who is in, noticing when a member disappears, repairing the chain, and handling the awkward situations when a member's site changes into something the others would rather not be linked to.
Rings collapsed because the chain breaks whenever a member vanishes, the repair work was unpaid and never-ending, the traversal often depended on one central script that could lapse, personal publishing moved onto platforms that decide what markup an author may use, and readers stopped arriving in a browsing mood.
Some current versions publish the member list as a public file, compute traversal without a central server, check member sites automatically, or publish a ceiling on membership. Others still run traversal through one central service, as the original ones did.
A ring has two things a register does not: every member consented, and it creates a community. A register has one thing a ring does not: a dated claim that somebody checked recently.
Sources
Each entry below is recorded in the source ledger under the identifier shown.
- S-002 · Casey 1998 — Web rings: An alternative to search engines, Carol Casey, College & Research Libraries News, vol. 59 no. 10 (1998); the journal's article page gives the volume and year but not the month. https://crln.acrl.org/index.php/crlnews/article/view/22105/28059 Retrieved 2026-09-06. Supports: what the ringmaster did — creating and maintaining the ring homepage, designing the elements shown on member sites, recruiting member sites, and developing a policy controlling the content and quality of the ring as applications grew.
- S-010 · Pew link rot report — When Online Content Disappears, Athena Chapekis, Samuel Bestvater, Emma Remy and Gonzalo Rivero, Pew Research Center, 17 May 2024. https://www.pewresearch.org/data-labs/2024/05/17/when-online-content-disappears/ Retrieved 2026-09-06. Supports: the random sample of just under a million Common Crawl webpages at approximately 90,000 per year from 2013 through 2023, the finding that a quarter of webpages that existed between 2013 and 2023 are no longer accessible, and the finding that 38% of webpages that existed in 2013 are not available.
- S-034 · WordPress.com code documentation — Add Code to Your WordPress Site, WordPress.com Support. https://wordpress.com/support/code/ Retrieved 2026-09-06. Supports: the published list of HTML tags all WordPress.com sites can use, including anchors and images; that
embed,frame,iframe,form,input,object,textarea,script,styleandlinkrequire a plugin-enabled plan with activated hosting features; and that JavaScript may be used on sites with a plugin-enabled plan. - S-018 · XXIIVV webring documentation — Webring, XXIIVV, repository README. https://github.com/XXIIVV/webring Retrieved 2026-09-06. Supports: joining by editing
index.htmland opening a pull request; the hash-based circular linking; the stated aim of sharing hand-crafted diaries, wikis and portfolios; the ten-content-page, about-page and own-domain criteria; the two-year inactivity removal; the 256-entry ceiling; and the collaborator succession path. - S-019 · XXIIVV webring page — Webring, XXIIVV. https://webring.xxiivv.com/ Checked 2026-09-06. Supports: that the ring was loading on that date, and the count of 250 member entries in the ordered list on its page.
- S-008 · IndieWeb webring — An IndieWeb Webring, maintained by Marty McGuire. https://xn--sr8hvo.ws/ Checked 2026-09-06. Supports: that it was loading on that date, that it "automatically keeps track of (in)active member sites", that "your previous and next neighbors are active in the webring", and that it works with sites supporting IndieAuth or RelMeAuth.
- S-003 · Hotline Webring — Hotline Webring. https://hotlinewebring.club/ Checked 2026-09-06. Supports: that it was loading on that date, the
/nextand/previousjoining instructions and automatic addition on first click, the statement that the operators look at every site that joins, the reserved right of removal, and the count of 860 rows in its published member table.
Try the stumble. One at a time, in the ring tradition, with a check date attached.