THE LANGUAGE OF DATA | PART 2 OF 7
How the Data Travels: FTP, RETS, and the Web API
The industry has rebuilt the pipe twice. The current rebuild, from copying to asking, changes who controls the data.
By Saul Klein and John Reilly September 2026
A data feed is a pipe: the automated delivery of listing data from the MLS into other systems. This article, which stands alone, is about what the pipe has actually been made of, because the industry has rebuilt it twice, and the current rebuild is quietly the most important one.
The FTP era: trucks full of files. In the beginning, feeds were literal file transfers. The MLS exported the database into flat files overnight, parked them on an FTP server, and the vendor downloaded the whole load, plus a folder of photos, and rebuilt its own copy on its own machines. Every recipient held a full private copy of the market, refreshed once a day if things went well. Nothing was standardized: every MLS had its own file layouts and field names, so a vendor serving ten markets maintained ten different translators. Slow, fragile, and impossible to police once the files left the building.
The RETS era: a standard way to copy. RETS, the Real Estate Transaction Standard, arrived around the turn of the century and governed the industry for nearly two decades. RETS gave vendors a standardized way to query the MLS and pull updates, every fifteen minutes instead of once a night, using common protocols instead of custom file layouts. It was a huge improvement, and it is important to see what it did not change: RETS was still a copying technology. The vendor still downloaded the data, still rebuilt a duplicate database on its own servers, and still served consumers from its copy, not from the MLS. RETS has been formally retired as a standard for years now, though pockets of it survive in practice, the way fax machines survive.
The Web API era: asking instead of copying. The current standard, the RESO Web API, is a different idea entirely. Instead of hauling the database home, an application can ask the MLS a question in real time, receive exactly the records it needs, in standardized fields defined by the RESO Data Dictionary, and display the answer. The pipe became a conversation. Modern feeds run over the Web API whether the recipient replicates or not, and every serious MLS now delivers through it.
Replication versus access: the distinction underneath the technology. Here is why the plumbing history matters to your rights and your business. Replication means the recipient copies the data and maintains its own duplicate database: the FTP model, the RETS model, and still the model behind most large sites today, because a portal serving millions of visitors needs the data on its own hardware. Access means the recipient queries the authoritative database on demand and keeps nothing, or nearly nothing: the data stays home, and every request passes through the front door where credentials are checked.
Under replication, enforcing a data license is an honor system. Once a copy exists on someone else’s servers, the license is a promise backed by audits and lawsuits, because the MLS cannot see what happens to the copy. Under access, the license becomes enforceable in real time: the MLS can meter every query, log every use, and cut off a credential the moment a violation appears. This is why the transport shift and the licensing shift now reshaping the industry are the same story. The industry spent thirty years handing out copies and then wondering where the data went. The Web API era makes it technically possible, for the first time, to grant use without surrendering possession, and every new rule you read, in California, in Texas, and soon everywhere, is being written by people who understand that difference.
This article stands on its own and is also part of the practitioner series “The Language of Data”. Copyright 2026 Saul Klein and John Reilly. All rights reserved. Commentary and general information, not legal advice.
