On Thu, Aug 11, 2016 at 1:12 PM, Aaron Bartell <aaronbartell@xxxxxxxxx> wrote:
I'd love feedback on things that you feel I don't explain well. I believe
"containers" play an important role in the future of running PASE workloads
on IBM i, and right now chroot is the best one we have (that I am aware
I get the broad concept of containers. I just meant that I don't know
specifically how chroot plays out on the i (surely there are some
quirks that come from that). I also don't know the ramifications even
on a mainstream Unix or Linux platform of what happens when you're
running in a chroot environment versus not. Ideally none, I am sure.
But we all know real life doesn't always match up with the ideal.
Btw, chroot (aka ibmichroot) is delivered with 5733OPS opt 3. IBM
named opt 3 "GCC" which is somewhat a misnomer.
IBM is famous for not being great at naming.
In short, 5733OPS opt 3
gives you ease of chroot creation via shell scripts, and ease of installing
perzl.org rpms via shell scripts.
For me, the holy grail is being able to issue pip install commands and
have them work even when the package being installed has parts that
are implemented in C and thus must be compiled for the target
platform. And that compilation should be transparent to whoever is
issuing the pip command. When I first heard about Option 3, I was
hoping that it would serve that purpose, but I'm increasingly getting
the sense that it doesn't, at least not yet; and also the sense that
this is an extremely tough nut to crack.
Kevin Adler has spoken a few times about this and related topics. I
know he would like this to work, and more generally for anything in
PASE, not just Python.