Summary

  • The 25 August 1991 comp.os.minix post asked MINIX users what they wanted; it was not a release announcement and did not name the project Linux.
  • Torvalds’s retrospective and the v0.01 notes distinguish a limited September source snapshot from the 5 October 0.02 announcement for wider FTP distribution.

A post can be a beginning without being a release

The sentence remembered from August 1991 sounds like a launch because it is modest: a student describes a small project and asks for advice. But the post does not announce a finished operating system. Its subject is “What would you like to see most in minix?” and its audience is a newsgroup of people who already use MINIX. Linus Torvalds says he is building something for 386 and 486 AT-compatible machines, has begun months earlier, and has ported Bash and GCC. He asks what features readers want, while making no promise to implement them.

The date matters. Torvalds’s own 1992 collection places that question on 25 August and says that the 0.01 sources were made available around mid-September. It also calls 0.01 a token gesture to one interested correspondent, not a public release he was proud to announce. The version-control record supplies a more precise artifact: the v0.01 tag is dated 17 September 1991. A message, a source tree, and a release to a broad audience are therefore three different milestones.

The network preceded the download

The route began earlier than the August post. On 3 July, Torvalds asked the same newsgroup for an electronic copy of the POSIX specification and preferred FTP sites. In his later recollection, that question drew unexpected replies, led to correspondence with a reader he calls arl, and helped produce a Linux directory on nic.funet.fi. This is a participant’s memory written a year later, not a traffic log, but it describes how a request for information could find people able to provide a place for code.

The August message then used the group as a place to solicit preferences from MINIX users. The thread connected a developing kernel to people who understood the environment in which it might run. That does not tell us how many readers saw it or which suggestions shaped the source. It shows a concrete form of networked development: ask a technically relevant group, continue some exchanges by email, and make files reachable from an FTP host.

What 0.01 could do

The v0.01 notes call the system immature and narrowly tied to 386 AT hardware. They say there were no public kernel binaries, that readers had to configure and compile the source, and that MINIX was still needed to bootstrap or set up the environment. They also say the source had produced a running kernel on two machines. Torvalds’s later description that 0.01 was “not actually runnable” appears to use a different threshold: a source snapshot that could run only with substantial existing tools and setup was not yet a usable standalone system.

That distinction is more useful than forcing the records into one verdict. The kernel could execute in a constrained development environment while still failing the practical test of an independent installation. A source archive made inspection possible; it did not remove the compiler, bootstrapping, hardware, or library dependencies.

October made the invitation wider

On 5 October Torvalds posted again, this time offering version 0.02 for wider distribution. He pointed readers to nic.funet.fi:/pub/OS/Linux, described full kernel source and a few binaries, and invited comments and contributions of freely distributable utilities. The same post warned that MINIX-386 was still needed to compile and set up the system, and that the library was not fully distributable. This is a release announcement, but it is still an invitation to technically capable users, not a claim of a complete or standalone operating system.

The change between August and October was not simply a higher version number. The first post asked what an intended audience wanted. The October post connected that audience to a named source directory and a more capable artifact. Usenet made the question visible; FTP made the files retrievable. Neither channel, by itself, proves adoption. The available record does not count downloads, successful installations, or later contributors.

RFC 977, written in 1986, describes Internet news distribution through NNTP and also distinguishes the Usenet news system from mailing lists and older host-to-host feeds such as UUCP. It helps explain the kind of infrastructure in which a news article could circulate. It does not trace the path of this particular post through specific servers. The careful history is therefore a sequence of documented offers, not a legend in which one post instantly created a complete Linux system.

Sources