On 2025/06/09 20:52, sirjofri via 9fans wrote:
Regarding everything else you wrote, the idea of Plan 9 is quite simple,
  I think: every resource can be abstracted as a filesystem, and thus be
presented in a namespace. The challenge is to actually do the work of
designing that abstraction.

The math is missing. But as Brian Stuart points out, it's early days in a "science" (and it is indeed a science) that effectively encompasses all other sciences in that it is a model of anything that can be expressed as a mathematical model.

Whether a filesystem is as "Turing complete" as the general purpose computer you carry in your pocket deserves being explored, but as a first approximation, Plan 9 has proved quite adequate.

That said, it's not really the point, but your statement correctly makes the case that the filesystem has proved a fine tool for most Computer Science applications; in my opinion it may fall short in fields that are yet to be assimilated under the Turing complete umbrella, but that is a bridge to cross when the need arises. I suppose I ought to mention "ioctl" and "Unix Sockets" as ridiculous alternatives.

My own beef, on the other hand, is that there are quite a few, much less general than "filesystem uber alles" within Plan 9 and certainly less pervasive, that still reflect the Plan 9 ethos. In that respect, I think it is important not to give the filesystem all the authority, but rather find, deeper than the filesystem, the axioms that Plan 9 is not willing to reject. This is purely a philosophical issue, but it is concerning that it may be overlooked simply because the filesystem is such a bright light and the others are less so.

Just because an example has just come to mind, which is in fact related to the filesystem: the proto files. I don't think the idea even has much of a name, but I have often thought that GIT-like utilities would do much better cataloguing their objects under a proto dataset than rely on the directory structure inherited from Unix. The fact that the proto library is incomplete - it typically does not provide a tool to manage "symbolic links", for example - is what jumps out at me. Something that just does not get attention when it is likely to find much greater usefulness, because the filesystem hammer is so much better at knocking nails.

Again, I have given this only rare thought, so I may well be barking up the wrong tree. Feel free to contradicting me with whatever logic I may have missed.

Lucio.

------------------------------------------
9fans: 9fans
Permalink: 
https://9fans.topicbox.com/groups/9fans/Tf84d656c78bbda91-M43697dee5613099f45fa0de1
Delivery options: https://9fans.topicbox.com/groups/9fans/subscription

Reply via email to