Question on multiple Python users in one application

[Gregory Ewing]

Loren Wilton wrote:
> I've read that Python supports 'threads', and I'd assumed (maybe 
> incorrectly) that these were somewhat separate environments that could 
> be operating concurrently (modulo the GC lock).

Not really. Python threads are just a thin wrapper around
OS threads, and don't provide any more isolation than the
OS's native threads do. All the threads in a Python process
share one set of imported modules, and therefore one set
of global data.

For example, there will only be one sys module, and thus
things like sys.stdin and sys.stdout will be shared between
threads. If one user pointed his sys.stdin somewhere else,
it would affect all the others.

There is *some* support in CPython for multiple interpeters
within a process, but I don't know how much isolation that
provides, or whether it has even kept up enough with all the
changes over the years to still be usable. I know that
subinterpreters are not fully isolated, e.g. built-in
constants and type objects are shared.

There is also the GIL to consider, which will prevent more
than one thread from running Python code at the same time.

Does Python really need access to data in the VM's memory,
or just to data in its disk files, etc? If the latter,
you might not need to run Python in the same process as
the VM as long as you can arrange some kind of I/O bridge.

Does the OS running on the VM have any notion of a remote
file access protocol a la NFS that an external Python
process could use to access data?

-- 
Greg