Hosting Through the Years / 1994 to 1996: Encryption arrives

1994 to 1996: Encryption arrives

HISTORY

3 min read · 751 words

As the first online shops appeared, people wanted to send card numbers without them being read in transit. Netscape introduced SSL in the mid-1990s as a way to encrypt web traffic, and the padlock became part of the browser. Early versions had flaws, and the protocol was gradually replaced by TLS, but the pattern was set: the browser verifies the server, then encrypts the conversation.

At first, HTTPS was used only for checkout pages. Certificates were expensive and had to be bought by hand, which is why it took twenty years for encryption to become the default.

The reasons it took so long are a good tour of how a technical standard meets money, law and habit.

The problem with plain HTTP

An ordinary web request crosses several networks on its way: the home or office router, the internet provider, a series of carriers, and the hosting company's own network. Over plain HTTP, every one of those can read the request and the reply, and anyone who controls one of them can change the content. For a page of opening hours that is a nuisance. For a form carrying a card number or a password, it is the whole risk.

The web needed two things at once: privacy, so that onlookers could not read the traffic, and identity, so that the browser could be reasonably sure it was talking to the real shop rather than a copy. Encryption alone does not give you the second; you can have a private conversation with an impostor.

SSL arrives, and the padlock with it

Netscape designed SSL for its browser and server software. The first version was never released publicly. Version 2.0 shipped in 1995, and a substantially redesigned version 3.0 followed in 1996 after security weaknesses were found in the earlier one. The browser showed a padlock or similar mark once a secure connection was in place, which gave ordinary shoppers a visible sign to look for.

Under the hood the handshake did three jobs. The server presented a certificate, vouching for its name. The two sides agreed on keys. Everything after that was scrambled with a fast shared-key cipher.

BrowserServer1. Hello, here is what I support2. Certificate for the site nameBrowser checks the certificate3. Key material, agree a shared key4. Encrypted traffic both ways
A simplified view: verify the server first, then encrypt. Modern TLS is shorter, but the order of ideas is the same.
1995SSL 2.01996SSL 3.01999TLS 1.02008TLS 1.22018TLS 1.3Spacing is not to scale. TLS 1.2 and 1.3 are the versions in use today.
Versions of the protocol, from Netscape's SSL to today's TLS 1.3.

Why it stayed on the checkout page

For the first twenty years HTTPS was treated as a premium feature, and there were reasons. Certificates came from commercial authorities, who checked an applicant and charged a fee, usually renewed yearly. Getting one meant generating a key and a request file on the server, pasting it into a web form, waiting, and installing the result by hand. Mistakes in that chain produced outages.

Encrypting everything also cost server capacity, which mattered on the hardware of the period. And in the early days a certificate generally wanted a dedicated IP address, since the server had to present the right certificate before it knew which name the browser was asking for. Extra addresses meant extra cost. A shop might reasonably decide to encrypt only the pages that handled payment.

SSL gives way to TLS

The IETF standardised the protocol as TLS 1.0 in 1999, which was close to SSL 3.0 with a new name. Later versions followed, with TLS 1.2 in 2008 and TLS 1.3 in 2018. The old SSL versions were retired as weaknesses became practical to exploit; SSL 3.0 was formally prohibited in 2015, and TLS 1.0 and 1.1 were deprecated in 2021. Many people still say "SSL certificate", and that habit is harmless, but the protocol on the wire today is TLS.

Another legacy of the period was the export rules of the United States, which restricted strong cryptography in software sold abroad. Weak "export grade" cipher options stayed in some servers for years and were later used in attacks, a reminder that a feature can outlive the reason for it.

What changed in the end

Three things moved at once. Free, automated certificates removed the fee and the manual work. Browsers began to mark plain HTTP pages as not secure, which turned HTTPS from an extra into an expectation. And server software learned to pick the right certificate by name, so a single address could serve many HTTPS sites. Today a host issues and renews certificates in the background, and the interesting failures are the ones where renewal fails without anyone noticing, which is a matter for the troubleshooting guide.

Previous1993 to 1996: The first hostsNext1995 to 1999: Dynamic sites and the domain rush

More from Hosting Through the Years

History

2006 to 2012: The cloud arrives

In 2006 Amazon launched S3 for storage and EC2 for computing, letting anyone rent capacity by the hour with a...

History

2000 to 2005: Dot-com aftermath and the rise of blogs

The bursting of the dot-com bubble thinned out the first wave of hosting companies, and the survivors...

History

2018: Privacy law reshapes the registry

When the European GDPR came into force in May 2018, the public WHOIS system, which listed names, addresses...