[atoav]

File system is the wrong word. What they should have used is file format.

It is not wrong that you can have a file (bits and bytes encoded in the shape described by a file format) on some remote point. If you have an index of those files where you can programmatically choose between multiple files that could even pass as a crude "file system". But I doubt this is what they meant to refer to.

It is likelier they wrongly assumed a file system is the system in which a file is organized, where in fact they meant file format.

[T-A]

I think they were going for a reasonable analogy, especially when a stream is saved to disk to have its contents extracted: each channel contained in the stream can then be thought of as a separate file, not unlike files in a zipped directory.

[LocalH]

I'd almost call it an "interleaved archive format" for a layman, but even archive isn't quite the right word here.

[atoav]

The word for this in media is container file format

[LocalH]

That's probably better, but also that's not much different on a conceptual level from a "filesystem". Both contain multiple files, and describe where in the overall image those files are. The difference is exactly in the interleaving - a multiplexed media container is somewhat constrained to a packet-based scenario, where individual packets of the constituent files are separated so that video, audio, and ancillary data all reach the decoder at roughly the same time. A pure filesystem is not constrained as such and can put the files anywhere within the container.

A filesystem stored in a monolithic file is not so constrained. I work with Harmonix games as a modder, and they use a bespoke format called "ARK", which is a two-part format. There exists one or more "ARK parts", which in implementation are virtually concatenated, and then the HDR lists a binary offset from the beginning of the first ARK part to tell the games where the individual files are. This could also be called a "container" conceptually. But none of such files are interleaved in any way except for the audio files, which use encrypted multichannel Ogg Vorbis streams plus a bespoke header to aid in seeking.

Plus, the context was explaining it to laypeople, where the specific jargon is less important than imparting knowledge and understanding.