----- Original Message -----
> > Storing user data under /run *is* controversial, and I object strongly
> > to it.
<snip>
> > /tmp exists specifically for this purpose, and while
> > you've pointed out that problems exist with /tmp, these are entirely
> > self-inflicted and are easily resolvable.
>
> /tmp exists as place where mkstemps() and mkdtemp() can be used on. This
> should be the only API for /tmp.
To use your terminology, "/tmp itself is an API". See POSIX:
> The following directory shall exist on conforming systems and shall be used
> as described:
>
> /tmp
> A directory made available for applications that need a place to create
> temporary files. Applications shall be allowed to create files in this
> directory, but shall not assume that such files are preserved between
> invocations of the application.
> Everything else is problematic for
> security reasons and that's why we should *never* place sockets there.
This statement is too general to be true (mktemp-like use of bind() and using
environment variables to point to sockets can be perfectly secure easily
enough), and even if it were true, it wouldn't automatically make /run/user the
generally accepted solution.
Mirek
_______________________________________________
fhs-discuss mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/fhs-discuss
