[MIR] python-sysv-ipc

[James Page]

Avaliability: In universe
Rationale: python-tooz switched from posix-ipc to sysv-ipc this cycle; used by OpenStack
Security: No CVE's found.
Quality assurance: No unit test suite upstream.
Dependencies: All in main
Standards compliance: OK
Maintenance: Server Team
Rationale: (From Upstream)

"Remove the usage of lockutils (and surronding oslo modules)
and keep it much simpler by just using the sysv_ipc module
directly.

This adjustment reduces the complexity, test exposure and
API complexity of tooz; removing much of the complex functionality
that was pulled in when sucking in the lockutils module for its
simple (but still broken) posix_ipc logic."

[Michael Terry (mterry)]

What about the rest of the reverse-dependencies for python-posix-ipc?

* python-ceilometer
* python-glance
* python-heat
* python-ironic
* python-keystone
* python-nova
* python-pycadf

Are they planning to switch too? Ubuntu abhors avoidable duplication in main.

[Michael Terry (mterry)]

Packaging etc seems fine besides though.

[James Page (james-page)]

Michael

Note sure how much I can do about this divergence; tooz appears to have more cross platform requirements which are met better with sysv_ipc, whereas the rest of OpenStack does not have plans to make this move.

Ceilometer is the only project using this component, and its not using the IPC support AFAICT so I'm tempted to drop the BD and associated ipc tests - as this is a dependency for testing only in later versions of tooz.

[Michael Terry (mterry)]

James, I don't follow your second paragraph. Can you explain like I'm 5? :)

[James Page (james-page)]

Michael

Tooz presents an number of ways of coordinating workflow between processes - ipc is one option, and is the least likely in a ceilometer deployment where components will be distributed across a number of systems.

So although IPC support its interesting, its not really a core target for the primary driver for main inclusion of tooz, so skipping that part of the test suite probably hurts us least, whilst ensuring we don't have two implementations of python-*-ipc in Ubuntu main.

[Michael Terry (mterry)]

Ah, I get it. So basically you're suggesting we drop the need for this MIR to help avoid double implementations. I'm fine with that. It would still be an optional run-time dependency, just not a build-dep / tested? Sounds good.