The Linux operating system itself can go much lower than the usual desktop applications that run on it.. As stated at IBM's Embedded Linux page:
a Linux system can actually be adapted to work with as little as 256 KB ROM and 512 KB RAM.
If they didn't mind running telnet for their web browser and ed for their word processor (one at a time!), they might need only another 128 Kbytes of RAM. But this is for serious (and insanely impoverished) geeks. Beware that none of the usual command line shells, not even sash - Stand-alone shell will go quite that low. When I have to work down here, I write my own shells using a page or two of C code to fork and exec simple commands, one at a time, as read from the input.
Linux can also stretch the other way. I've run on it systems with 2048 CPUs and 2 terabytes of main memory. This wasn't a cluster; this was a single system image (SSI) multi-processor system.
> True, but not because of Linux. Rather, this is true because of the various applications, such as web browser or word processor, that a general purpose home user will want to run.
True, for the most part. The other part is feature set -- you can do a small, tight-featured web browser in about a quarter the size of today's heavily-featured browsers.
> The Linux operating system itself can go much lower than the usual desktop applications that run on it.. As stated at IBM's Embedded Linux page: a Linux system can actually be adapted to work with as little as 256 KB ROM and 512 KB RAM.
I think we're really discussing various definitions of "a Linux system". Linux is the kernel. The GNU tools and other OS components, plus a host of applications, make up what we loosely call "a Linux system".
> If they didn't mind running telnet for their web browser and ed for their word processor (one at a time!), they might need only another 128 Kbytes of RAM. But this is for serious (and insanely impoverished) geeks. Beware that none of the usual command line shells, not even sash - Stand-alone shell will go quite that low. When I have to work down here, I write my own shells using a page or two of C code to fork and exec simple commands, one at a time, as read from the input.
In the mid-80's, I ran a 32-bit desktop minicomputer (AT&T 3B2/300) with 2MB of RAM and a 30MB MFM hard drive that only had 10MB of system files on it. Yet, it ran Sys-5 Unix, and I had a full-featured WYSIWYG screen editor, word processor, communications software, email, a couple different shells, etc. and could support a couple of concurrent logged-in users. No heavy graphics, of course, all text.
When I was on it by myself, I could run in under 512KB of RAM, and as long as I wasn't paging too badly, it ran acceptably. And that was a full, legit, Unix. I'm sure Linux could do better than telnet and ed. Vi at least.
> Linux can also stretch the other way. I've run on it systems with 2048 CPUs and 2 terabytes of main memory. This wasn't a cluster; this was a single system image (SSI) multi-processor system.
Indeed, that's a major strength of Linux -- scalability.