1971年，IBM推出了第一个“存储磁盘”，即今天更为人所知的“软盘”。它是一个8英寸的软塑料磁盘，表面涂有磁性氧化铁。计算机数据被写入磁盘表面并从磁盘表面读取。第一张Shugart软盘存储了100kbs的数据。

“软盘”的昵称来源于磁盘的灵活性。软盘是一圈磁性材料，类似于盒式磁带等其他类型的录音带，其中磁盘的一面或两面用于录音。磁盘驱动器抓住软盘的中心，像在外壳内旋转唱片一样旋转。读/写磁头与磁带组上的磁头非常相似，它通过塑料外壳或信封中的开口与表面接触。

软盘因其便携性被认为是“计算机史”上的革命性设备，它提供了一种从计算机到计算机传输数据的新的、简单的物理手段。由艾伦·舒加特（Alan Shugart）领导的IBM工程师发明的第一批磁盘设计用于将微码加载到Merlin（IBM 3330）磁盘包文件（一种100 MB存储设备）的控制器中。因此，实际上，第一批软盘被用来填充另一种类型的数据存储设备。后来人们发现了软盘的其他用途，使它成为热门的新程序和文件存储介质。

5 1/4英寸软盘

1976年，Alan Shugart为Wang实验室开发了5 1/4“软磁盘驱动器和软盘。Wang希望使用更小的软盘和驱动器与台式计算机配合使用。到1978年，超过10家制造商生产了5 1/4”软磁盘驱动器，可存储高达1.2MB（兆字节）的数据。

关于5 1/4英寸软盘的一个有趣的故事是确定磁盘大小的方式。工程师Jim Adkisson和Don Massaro正在与Wang实验室的An Wang讨论尺寸。这三个人碰巧在一家酒吧​王向一张餐巾示意，并说“大约那个尺寸”，刚好是5 1/4英寸宽。

1981年，索尼推出了第一批3 1/2英寸软盘驱动器和软盘。这些软盘封装在硬塑料中，但名称不变。它们存储400kb的数据，后来是720K（双密度）和1.44MB（高密度）。

如今，可刻录CD/DVD、闪存驱动器和云驱动器已经取代软盘，成为将文件从一台计算机传输到另一台计算机的主要手段。

使用软盘

下面的采访是由Richard Mateosian完成的，他为第一批“floppies”开发了软盘操作系统。Mateosian目前是加利福尼亚州伯克利市IEEE Micro的评论编辑。

用他自己的话来说：

The disks were 8 inches in diameter and had a capacity of 200K. Since they were so large, we divided them into four partitions, each of which we regarded as a separate hardware device -- analogous to a cassette drive (our other main peripheral storage device). We used floppy disks and cassettes mostly as paper tape replacements, but we also appreciated and exploited the random access nature of disks.

Our operating system had a set of logical devices (source input, listing output, error output, binary output, etc.) and a mechanism for establishing a correspondence between these and the hardware devices. Our applications programs were versions of HP assemblers, compilers and so forth, modified (by us, with HP's blessing) to use our logical devices for their I/O functions.

The rest of the operating system was basically a command monitor. The commands had mainly to do with file manipulation. There were some conditional commands (like IF DISK) for use in batch files. The entire operating system and all of the application programs were in HP 2100 series assembly language.

The underlying system software, which we wrote from scratch, was interrupt driven, so we could support simultaneous I/O operations, such as keying in commands while the printer was running or typing ahead of the 10 character per second teletype. The structure of the software evolved from Gary Hornbuckle's 1968 paper "Multiprocessing Monitor for Small Machines" and from PDP8-based systems I worked on at Berkeley Scientific Laboratories (BSL) in the late 1960s. The work at BSL was largely inspired by the late Rudolph Langer, who improved significantly on Hornbuckle's model.