1996. Before a search box could read the whole web, being found meant a person read your site and decided which shelf it went on.
add.yahoo.com
The submission queue is closed. Press Open the queue above and the first site arrives.
Nothing is timed yet. The first two are walked through.
A trailing @ on a bold heading means the link leaves this branch of the tree entirely. This is the real Computers and Internet page, captured 20 December 1996. It lists 33 headings. Thirteen of them are @.
Semiconductors@ points at /Science/Engineering/Electrical_Engineering/Semiconductors/. The heading appears under Computers, the sites live under Science, and the count in italics belongs to the branch it points into, not to this page. A tree that needs this many shortcuts is admitting it is not a tree.
The Yesterweb link directory, a modern revival, describes its own ordering: Every time this page reloads, the order of websites are shuffled.
Recommendations decay and disappear completely after 30 days. Reload both of these and watch what each one does.
Loads 1
Votes compound, so position one keeps winning because it is already position one. Shuffling is not a worse ranking. It is a refusal to have one, and it is the only way a small list stays a list instead of becoming a leaderboard.
How to spot a directory.
Semiconductors@A trailing @ on a category name
The single most recognizable mark. It means this heading is a cross-reference into another branch, and the sites are not really here.
Software (3760)A count in italic parentheses
after the heading, totalling everything below it: Software (3760). A number nobody could keep accurate by hand, printed as though they could.
add.yahoo.com/bin/add?An "add your site" link
usually in the footer or the menu bar. On the 1996 homepage it reads "How to Include Your Site" and points at add.yahoo.com. Submission is a request, not an action.
about dmoz | add URL | feedback | link | editor loginAn editor login
The 1999 DMOZ nav row runs "about dmoz | add URL | feedback | link | editor login". That last one is the whole labor model sitting in the navigation.
<h3><a href="/">Top</a>:Computers and Internet</h3>A breadcrumb that is a heading, not a nav bar
<h3><a href="/">Top</a>:Computers and Internet</h3>. No separator graphics, no chevrons, just a colon.
<body>No stylesheet at all
The 1996 Yahoo! body tag is literally <body>. Every visual decision is structural: <center>, <ul>, <b>, <em>, and a bare <p> used as a spacer between list items.
A directory let you be found by subject rather than by phrase. You did not have to guess which words a stranger would type, because a person had already decided where you belonged and put you next to your neighbors. Browsing a branch showed you the whole of it, and the order was nobody's ranking.
What it traded was arithmetic. Every listing cost a human decision, and decisions do not scale with the web. The moment full-text search was good enough, the shelf collapsed into a single input box, and the question changed from where does this belong to what matches this string.
Somebody read your site and put it on a shelf, next to the others like it.