make relay_state in IDP response optional

[rhoerbe]

Required for redirecturl microservice because the redirect url handler does not have an opportunity to set the backend relay_state.

(Remark: IIUC satosa only sets a backend relaystate in the AuthnRequest to compare it with the Response, but does nothing else. In the context of SAMl I see no use for the relaystate in the backend)

[peppelinux]

Hi guys, I think that we should implement our ideas following the goals related to these. Code style and patterns are very important but they shouldn't block the availability of a feature in a usefull time.

We could do refactor and coding style purposes in a second time, we Just have to keep track of them in mailinglist or github.

[rhoerbe]

Travis is fine with py3.6, 3.7 and 3.8. Only py3.5 is failing - it should be removed.

[peppelinux]

Travis is fine with py3.6, 3.7 and 3.8. Only py3.5 is failing - it should be removed.

Important decision. Ubnt 16.04 and debian9 still have py35 as default

[ioparaskev]

Travis is fine with py3.6, 3.7 and 3.8. Only py3.5 is failing - it should be removed.

Can you explain why this error exists only in py3.5 and why py3.5 should be removed?

[peppelinux]

Travis is fine with py3.6, 3.7 and 3.8. Only py3.5 is failing - it should be removed.

Can you explain why this error exists only in py3.5 and why py3.5 should be removed?

See travis log, it seems that is something related to socket management through urllib3. Probably the code refactor touched some feature that's not available in py35. I should dug the code, but I think that this the problem

[peppelinux]

Bytheway removing py35 could be done in a future release, I think that 2021 should be the year to abadon py35

[ioparaskev]

Travis is fine with py3.6, 3.7 and 3.8. Only py3.5 is failing - it should be removed.

Can you explain why this error exists only in py3.5 and why py3.5 should be removed?

See travis log, it seems that is something related to socket management through urllib3. Probably the code refactor touched some feature that's not available in py35. I should dug the code, but I think that this the problem

There isn't any obvious python 3.5 breaking change in the PR and master branch currently is green for python 3.5. I'm saying that we (the author of the PR I would say) should investigate why this error happens in py3.5 (maybe it's a travis issue related to py35?), check if we can solve it and what compromise we might have to do to solve it, instead of suggesting immediately the removal of python 3.5.

[rhoerbe]

I agree. I forgot the debian dependency on py3.5

[rhoerbe]

May be that is a karmic hint that I should appreciate @ioparaskev 's style sugestion ;-)

[ioparaskev]

It seems there is an issue with travis and py3.5 with the requests package (worth investigating why these random failures happen). @c00kiemon5ter re-triggered the build and now everything passes :wink: