Hi Steve: Thanks for your great efforts for WSME replacement. I had a quick review for your code changes, see my comments below: Review Feedback Reason 704490 Has concern move wsme code to ironic, types.py 704487 Has concern move wsme code to ironic, node.py 704489 has concern move wsme code to ironic, args.py 704486 has concern move wsme code to ironic expose.py 703898 Both positive and concern Positive, better code structure than original code, concern, more difficult to locate WSME code 704488 has concern move wsme code to ironic 704485 has concern Add pecan code, pecan would be replaced by flask 703897 Positive Removed some WSME code with this change 703723 Positive Removed some WSME code with this change 703695 Positive Removed some WSME code with this change Firstly, I appreciated 3 changes which definately removed some wsme code, especially the change 703897. But I have some concerns for changes 704490, 704487, 704489, ..., these changes seemed move WSME code into ironic, that would make ironic code base become a bit large, from my personal view, to use python built-in feature or other 3-party lib to replace WSME function would be better. I like the way that code change 703897 did. Thanks Jerry 发自我的iPhone 在 2020年1月30日，上午8:53，Steve Baker <sbaker@redhat.com> 写道： I've put together a set of changes for removing WSME which involves copying just enough of it into ironic, and I think we need to have the conversation about whether this approach is desirable :) This git branch[1] finishes with WSME removed and existing tests passing. Here are some stats about lines of code (not including unit tests, calculated with cloc): 4500 wsme/wsme 6000 ironic/ironic/api master 7000 ironic/ironic/api story/1651346 In words, we need 1000 out of 4500 lines of WSME source in order to support 6000 lines of ironic specific API code. Switching to a replacement for WSME would likely touch a good proportion of that 6000 lines of ironic specific API code. If we eventually replace it with a new library I think it would be easier if the thing being replaced is inside the ironic source tree to allow for a gradual transition. So this approach could be an end in itself, or it could be a step towards the final goal. My strategy for copying in code was to pull in chunks of required logic while removing some unused features (like request pass-through and date & time type serialization). I also replaced py2/3 patterns with pure py3. There is likely further things which can be removed or refactored for simplicity, but what exists currently works. If there is enough positive feedback for this approach I'll start on unit test coverage for the new code. [1] https://review.opendev.org/#/q/topic:story/1651346