---
title: "Interlude. Before the Abstraction Existed"
url: "https://www.mikefluff.com/en/read/from-closures-to-autonomous-systems/interlude3"
locale: "en"
author: "Mikhail Savchenko (Mike Fluff)"
updated: "2026-11-18T12:24:57.879Z"
book: "https://www.mikefluff.com/en/books/from-closures-to-autonomous-systems"
access: "lead-in"
purchase: "https://www.mikefluff.com/en/books/from-closures-to-autonomous-systems"
alternates:
  - locale: "ru"
    url: "https://www.mikefluff.com/ru/read/from-closures-to-autonomous-systems/interlude3"
    markdown: "https://www.mikefluff.com/ru/read/from-closures-to-autonomous-systems/interlude3.md"
  - locale: "en"
    url: "https://www.mikefluff.com/en/read/from-closures-to-autonomous-systems/interlude3"
    markdown: "https://www.mikefluff.com/en/read/from-closures-to-autonomous-systems/interlude3.md"
  - locale: "pt"
    url: "https://www.mikefluff.com/pt/read/from-closures-to-autonomous-systems/interlude3"
    markdown: "https://www.mikefluff.com/pt/read/from-closures-to-autonomous-systems/interlude3.md"
---

# Interlude. Before the Abstraction Existed

*[JavaScript: From Closures to Autonomous Systems](https://www.mikefluff.com/en/books/from-closures-to-autonomous-systems)*

Windows NT had an assumption that was written down nowhere, because it seemed obvious: one person sits at the computer. The assumption was spread throughout the system. The graphical part of the kernel, the `win32k.sys` module, kept its data in a single instance per machine. Named objects through which programs coordinated with each other, mutexes and events, lived in one shared namespace. Programs put their settings in a shared registry branch and in INI files in the Windows directory, and that was reasonable as long as there was one user. The `csrss` process, which served the console and windows, was also one.

In the mid-1990s Citrix decided that one machine could serve many people at once, each in their own session, with their programs running on the server. To do that, it licensed the source code of NT 3.51 from Microsoft and in 1994 released WinFrame. The technology was called MultiWin, and to build it Citrix had to rework the kernel’s virtual memory manager and object manager. In May 1997 Microsoft and Citrix signed an agreement under which MultiWin would go into future versions of Windows NT. The project got the code name Hydra, the first beta was shown in November 1997, and on 16 June 1998 Windows NT Server 4.0 Terminal Server Edition was released to manufacturing as a separate build of the operating system.

* * *

In December 1998 Frank Kim described in Microsoft Systems Journal what had to change for this. A new kind of address space appeared, called session space: part of the kernel’s addresses became private to the group of processes of one user, and every user had there their own copy of the data of `win32k.sys`, the printer drivers and the display drivers. The object manager was reworked so thoroughly that it got a new name, the Multi-User Object Manager. Each user got their own namespace, so that two people could create a mutex with the same name without getting in each other’s way, and for objects that had to be shared, the `Global\` prefix was introduced. A separate copy of `csrss` was started for each session.

---

The rest of this chapter is free on the site once the reader confirms an email address: https://www.mikefluff.com/en/read/from-closures-to-autonomous-systems/interlude3
