Archives that vanish, and how to keep one

Webcomic archive preservation sounds like a librarian’s concern until you read the list of what is gone. The web was supposed to be the channel that never went out of print. It turned out to be the only comics channel where the entire back catalogue can stop existing on a Tuesday because a company revised its plans.

Print decays slowly and survives by accident, in boxes, in shops, in libraries that nobody consulted about it. A hosted archive is one copy in one place, and there are no accidents available.

The list

Zuda Comics was DC’s webcomics imprint. Its first competition line up went live on 30 October 2007, the competition format was dropped in April 2010, and the site closed that July. A few flagship titles moved to a commercial digital service. The rest of the catalogue, running strips and cancelled ones alike, stopped being available where it had lived.

Modern Tales, which Joey Manley launched on 2 March 2002, was one of the first serious attempts at paid subscription webcomics. It shut down in 2012. Webcomics Nation, which Manley launched on 29 July 2005, closed with the rest of his sites in April 2013. Manley died in November of that year, after the closures rather than before them, which is worth stating because the order gets reversed in the retelling.

Inkblazers, previously MangaMagazine, closed on 1 February 2015 and told its creators the audience had never been large enough to pay for the service.

Smack Jeeves is the best documented case and the most useful one. A redesign in December 2019 replaced the site’s page customisation with a scrolling format and drove a large part of its community off. Closure was announced on 1 November 2020 and the site went dark on 31 December 2020. In those two months volunteers uploaded 34,757 of roughly 36,100 hosted comics to the Internet Archive. That is an extraordinary rescue, done by unpaid people in a hurry, and it still left more than a thousand comics that nobody got to.

Comic Genesis, the free hosting arm that grew out of Keenspot in 2000, never announced anything. It simply stopped working properly over a period of years. That is the other way an archive ends: not a shutdown notice but a slow failure that nobody is being paid to repair.

The Duck Webcomics, launched in 2002 by Dylan Squires and later owned by Platinum Studios, lost site data in a server failure at the end of 2005 and was rebuilt the following January. Accounts differ on how much came back, which is itself the lesson. Nobody keeps good records of what a crash took.

The general case predates webcomics. Yahoo closed the American GeoCities on 26 October 2009. Archive Team and the Internet Archive crawled it for months beforehand and released roughly 900 gigabytes covering tens of millions of pages, and that effort described itself as capturing a significant percentage rather than all of it.

Policy closes archives faster than bankruptcy

A shutdown is impersonal and usually announced. A policy change is selective and arrives with almost no notice.

Tumblr announced an adult content ban on 3 December 2018 and enforced it from the 17th. Two weeks. Automated classification caught a great deal that the written policy did not describe, and artists who had used the site as their archive for years did not find their work moved, they found it hidden. That is the mechanism in comics censorship after the Code applied to storage rather than to sales, and it lands hardest on exactly the work that has the fewest alternative homes.

What the Wayback Machine can and cannot do

It is the safety net and it has holes worth knowing about.

The Internet Archive was founded in 1996 and the Wayback Machine launched publicly on 24 October 2001. For its first sixteen years it honoured robots.txt strictly, which made archives retroactively erasable: a host that later added a restrictive file could black out captures taken years earlier, when nothing had been restricted. The Archive announced in April 2017 that it would generally stop treating robots.txt as an instruction about archiving. That improved everything after 2017 and repaired none of the gaps before it.

The second problem is specific to comics. An archive page whose navigation runs in JavaScript and fetches images from the original server does not replay once that server is gone. The crawler captured a shell. Sites made of plain pages with plain links between them archive well. Sites with clever galleries archive badly, and image heavy sites are exactly the ones most likely to have clever galleries.

Institutions have begun collecting on purpose. The Library of Congress launched a Webcomics Web Archive in June 2017 through its American Folklife Center, covering around fifty sites. That is real and welcome, and it is fifty sites.

The working version

None of this needs a budget. It needs treating the archive as the asset rather than as a byproduct of publishing.

Own the domain. Everything else is replaceable and the address is not. A domain can be pointed at a new host in an afternoon, so inbound links, bookmarks and accumulated search results survive the move. That argument opens where adult comics can live and it is the highest value item here by a distance.

Keep the masters off the web host. Full resolution layered files, on your own drives, with one copy that is not on the same drive and preferably not in the same building. What sits on a website is a compressed derivative of the work, not the work.

Publish plain. Stable readable addresses that never change, one page per instalment, ordinary links, and an index that lists everything. This makes the site archivable by anyone: the Internet Archive, a reader with a downloader, and you.

Mirror it. A second copy on unrelated infrastructure, refreshed occasionally rather than continuously, is enough to survive most failures. Two services owned by the same company count as one copy.

Export the list. Mailing list, subscribers, whatever exists. Followers belong to the platform and do not transfer.

Request captures yourself. Anyone can ask the Wayback Machine to archive a page. Doing it at the end of each chapter costs nothing and puts a third party copy outside your own control, which is the only kind that survives your own mistakes.

And print something. Forty copies of a book in forty houses is a distributed backup with no operator, no terms of service and nobody to change their mind. That is the oldest lesson in this medium arriving from a new direction, and it is why the platform era keeps ending the same way.