> So what you're saying is.... newsflash, the contents of a directory is
> the directory entries.

Well... This just shows you aren't familiar with the subject. There are
different ways to implement hierarchical filesystems. Take, for instance,
Ceph filesystem. It's built on top of an object store. It doesn't need
i-node per directory as the whole notion of directory is different: it's
just a common prefix shared by multiple objects. Directories aren't objects
in such a filesystem and don't have their own content.

I didn't have the time to test what happens to read-only directories in
Ceph, but wouldn't be surprised if the concept of read-only directory
doesn't exist there at all.

Furthermore, implementing directories as i-nodes is one of the valid
choices. It has cons and pros. Ultimately, networking filesystem would
gravitate towards storing directory data in a centralised way while
local-with-compute filesystem would probably prefer a distributed way of
storing directory data.

You are precisely the illustration of my previous description of a
developer who argues wrongly in defense of a selected behaviour based on
their particular familiarity with implementation details, while lacking
larger conceptual picture of the problem
-- 
https://mail.python.org/mailman3//lists/python-list.python.org

Reply via email to