[ovs-discuss] Why ovn's code structure is getting messy

[gongming chen]

To be honest, I think the direction of the ovn community is not towards a stable and clear structure. Especially in the patch integration in the past year, the new patch does not take too much into consideration for compatibility. Many of them are completely logical changes, so the cost of learning and following the ovn community is greatly increased. And many logical changes are completely unnecessary.
For example, ofctrl_dup_flow--->ovn_flow_dup, when looking at the new github code, I found that ofctrl_dup_flow was not found, and the new function did not change anything except the name. I think this change is completely unnecessary, except Will increase the time cost of reading new code.
Complexity such as (no other maliciousness), Incremental processing for flow installation by tracking:
Compared with the small benefits it brings (ofctl put itself is not a big bottleneck), I think the complexity it brings, and the changes brought by the code structure (the change of the existing version is a big cost) Is bigger. This is not conducive to the long-term stable development of ovn, because the code structure of ovn is becoming increasingly unclear.
As a general community participant, with long-term ovn production environment operating experience, I think the big bottleneck of ovn now is stability and structural clarity (for custom development). I hope ovn has clear code (like the code structure of the kernel), which can reduce the time cost of using it. I hope ovn is small, clear, stable, and reliable. Some personal suggestions, and finally hope that the ovn community will get better and better.
Best wishes.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openvswitch.org/pipermail/ovs-discuss/attachments/20210202/c5127cb3/attachment.html>