If you spend any time digging through the remains of 2000s webcomic sites, one pattern turns up again and again, on sites with no connection to each other: a folder per strip, a folder per issue inside it, and a numbered file per page inside that. Nobody standardised it. It emerged independently, over and over, because it was the obvious thing to do with a directory tree and a stack of scans — and because it made the next page's filename predictable enough to type by hand.
The canonical shape
The dominant form was three segments deep, zero-padded to two digits at each numbered level, with a static extension:
/lanterns/issue03/page07.htm— series, issue, page./lanterns/issue03/page07.html— the same thing on a site that had settled on the longer extension./lanterns/ch02/p14.htm— the abbreviated variant, common where the strip thought in chapters.
Zero padding was not a nicety. Without it, a directory listing sorted page1, page10, page11, page2, which made the coordinator's own file management miserable. Two digits solved it up to ninety-nine pages, which felt like plenty in 2004 and turned out not to be for anything that ran a decade.
Cover pages at page00
Once you have page01, a cover needs somewhere to go, and the elegant answer is page00. Sites did this constantly. It is worth knowing because it produces a small archaeological tell: an inbound link pointing at page00 is almost always someone linking to an issue as a whole, whereas a link to page01 is someone linking to where the story starts. Directories tended to pick one convention and apply it to every listing, which is why a whole cluster of links from a single referrer will often share the same page number.
Other sites used cover.htm, coverindex.htm, or an index.htm inside the issue folder that acted as a cover and a jump list at once. All three coexist in surviving link data, sometimes on the same site in different years.
Arc renumbering, the quiet archive-killer
Here is where it goes wrong. A long-running strip accumulates pages continuously, but a reader wants to enter at a story arc. Sooner or later the creator reorganises: what was a flat run of 200 pages becomes seven named arcs, and pages get renumbered within each arc. Or the reverse — named arc folders get a numeric prefix so they sort correctly in a listing, and /strip/darkwater/ becomes /strip/03-darkwater/.
Each of those reorganisations is individually sensible. Cumulatively they are lethal. It is entirely normal to find a strip that ran from 2004 to 2019 having used four mutually incompatible URL schemes over its life — an early per-episode scheme, a named-arc scheme, a numbered-arc scheme, and a final flat scheme — with each migration silently orphaning every external link made under the previous one. No redirects were left behind, because leaving redirects behind requires either server configuration the hobbyist did not have access to or a stack of stub pages nobody wanted to maintain.
This is the single biggest reason a directory entry from 2008 pointing at a strip that ran until 2019 is dead even though the site outlived it by a decade. The site did not go away. The address did.
Spaces, capitals and the case-sensitivity trap
Hand-built sites had hand-made folder names, and they show it. Within one site you will find lowercase single words, MixedCase compounds, a folder named after a person, a misspelling nobody ever fixed, and — reliably, at least once — a directory name containing literal spaces.
Spaces are the worst of these. A path with spaces gets percent-encoded by some clients and left literal by others, so the same page acquires two different addresses in the wild depending on who linked it. Combine that with inconsistent capitalisation and a case-sensitive server and one page can have four live variants, of which a later host migration may preserve only one.
The practical lesson has not changed: lowercase, hyphens, no spaces, no capitals, and never put a person's name in a path you might one day want to detach from that person.
The .htm and .html twins
Extension drift is the most common breakage of all, and the most avoidable. Early sites overwhelmingly used the three-letter extension — a habit inherited from the filesystem limits of earlier desktop software — and many later switched to the four-letter form when they rebuilt, or when a new host's default index file changed.
The switch rarely happened cleanly. What you typically find is a section with both index.htm and index.html having returned a valid page in different years, each cited externally by whoever was linking at the time. Both addresses are real history. Only one of them can be the canonical page on a rebuild, which means the other needs a permanent redirect if the citation is to survive — and a 301 rather than a 302, because a temporary redirect tells search engines to keep the old address indexed and consolidates nothing.
What a durable scheme looks like
Very little of what follows was available as advice in 2004, but all of it was available as principle — the W3C was arguing for stable URIs from 1998.
- Pick an extension and never change it. Or serve extensionless paths and let the server resolve them, which removes the decision entirely.
- Number pages sequentially for the whole strip, not per arc. Arcs are a navigation concept; let them be a page that links into the sequence rather than a directory that owns it.
- Pad generously. Three digits costs nothing and buys a thousand pages.
- Keep names free of people, dates and marketing. Anything that can change will.
- When you must move, redirect permanently, and keep the redirects forever. A redirect map is the cheapest artefact in web publishing and the only one that preserves other people's links.
None of this saves an archive whose images have gone missing, which is a different failure — see link rot and archive decay — and none of it helps if the domain lapses. But it is the difference between an archive that can be moved and one that can only be abandoned. The site anatomy page covers why so few of these pages carry any text to identify themselves with once the addresses stop meaning anything.