Messaging platforms timeline: four models, no dates

The messaging platforms timeline is not a parade of applications. It is the history of one question: what is your address, and who is allowed to write to it?

Every large shift in the category changed the answer, and each answer brought a matching set of problems. Read the sequence that way and a great deal that looks like fashion turns out to be consequence.

What to take away

Four addressing models, in order

The named platforms line up with those models. Email, XMPP (Jabber), and IRC belong to the open address. AIM, ICQ, MSN Messenger, Slack, and Discord belong to the service account. SMS, MMS, RCS, WhatsApp, Telegram, and Signal belong to the telephone number. WeChat, iMessage, Facebook Messenger, and Snapchat belong to the application account.

The open address. The first durable model gave you an address that anyone could write to from anywhere, without a shared provider and without permission. Delivery was store and forward, in the shape SMTP still uses: your message waited until the other end collected it. Anyone could run the infrastructure, so nobody could be excluded from participating, and the address was yours to publish.

The service account. Then messaging moved inside individual services. Your address became a name that only existed within one product, and reaching you required being on the same product. This bought presence (you could see who was available, which the open model could not do) and real-time delivery, and it introduced the idea that a conversation is a thing that happens in an application rather than between two addresses.

The telephone number. Mobile devices made the number the address. This solved discovery completely: everybody already had your number, so a new application could populate your contact list on the first launch by reading the address book. It also welded your reachability to a telecoms account, tied it to your legal identity in most places, and handed a permanent route to anyone who had ever been given the number.

The application as address. The current arrangement in most of the category. What you are reachable at is an account inside one operator's system, discovered through a contact list, protected by that operator's rules, and meaningful nowhere else. The address does not exist outside the software.

What flipped at each step

Read the bottom two rows together. The arrangements that are pleasant to be inside are the ones that are expensive to leave, and that is not a coincidence: the wall that keeps strangers out also keeps you in.

Spam is the engine of the whole sequence

Every step in that table is partly a response to unwanted contact, and the response shapes everything else.

An open address space has no way to make sending expensive, so defense has to happen at the receiving end: filtering, reputation, and lists of who is trusted. That defense works reasonably well and it has a permanent cost, which is that legitimate messages are sometimes silently discarded and nobody can be sure which.

A closed system makes sending expensive by making membership a prerequisite. Defense moves to the door: account creation limits, rate limits, and the ability to remove a sender entirely. Spam inside is far lower. The price is that the operator now decides who can speak, a much larger power than filtering ever was.

Once that power exists it has to be exercised by someone, on a schedule, against a written standard. That is the same problem that turns up wherever people gather in numbers, and it is set out under community platforms.

The number-as-address model got the worst of both. It is effectively open, because anyone who has the number can reach it, and the identity attached is unusually valuable, which is what makes it worth attacking.

None of these is a solved problem. They are three different placements of the same cost.

Presence, receipts, and other social machinery

Two features arrived with the service-account era and changed conversation more than any protocol did.

Presence told you whether someone was there. It converted messaging from correspondence into something closer to a room, and it created an obligation that had not existed: if the system says you are available, not answering becomes a choice that other people can observe.

Delivery and read receipts did the same thing to the individual message. The information is genuinely useful and it removes an excuse that people relied on. Where receipts cannot be turned off, or can only be turned off in both directions, the setting is a small piece of social policy written by the operator.

Both are worth noticing because neither is technically necessary. They exist because they increase engagement, and they persist because the norms they created are now expected. Whether either is on by default in your own account is one of the things you can settle with a five minute test rather than by reading about it.

The pendulum, and where it sits

Interoperation keeps coming back. It returns when a walled service grows large enough that being outside it is a real disadvantage. It also returns when running several applications for several groups becomes tiring enough that someone builds a bridge.

The European Union's Digital Markets Act is the clearest current rule pushing large messaging services toward interoperability. Matrix is a protocol built around that idea.

It gets rolled back when a bridge becomes a route for abuse, or when it stops being commercially attractive to the operator with the larger audience.

The trade is stable even though the outcome is not. Openness lowers the cost of leaving and raises the cost of defense; closure does the reverse. Systems designed to keep interoperability as a permanent property rather than a temporary concession are covered separately under decentralized networks, because their difficulties are different ones.

For a reader deciding what to use, the sequence matters in one practical way. Ask which addressing model you are inside, and the rest of the service's behavior becomes predictable: how you will be found, who can reach you, what leaving costs, and who decides those things. The messaging platforms overview sets out the questions that follow from the answer.

Common questions

Was the open model better?

It was better on portability and worse on almost everything users notice daily. It also carried a filtering problem that has never been solved, only managed. Preferring it is a defensible position, but it means preferring a set of costs, not avoiding them.

Why did the phone number win, given its problems?

Because it solved the coldest start in the category. A new messaging application with an empty contact list is useless, and reading the address book made it useful in one second. That advantage was large enough to outweigh every objection, and the objections have not gone away.

Will interoperation between the large services actually arrive?

Parts of it appear periodically, and the sticking points are not technical. Reputation, abuse handling, and who is accountable when a message crosses a boundary are the questions that decide whether a bridge lasts, and each one is harder than the message format ever was.

What does this history tell me to do now?

Know what your address is and what would happen if you lost it. If your reachability depends on a telecoms account or an application account, write down how the people who need you would find you if it vanished tomorrow. That plan takes ten minutes and is the only part of this history that acts on your behalf.