OpenVPN 3 fork – merge to upstream?

[benfitita 39]

@Staff In your Black Friday post you mention:
 

Quote

AirVPN is the only VPN provider which is actively developing OpenVPN 3 library with a fork that's currently 190 commits ahead of OpenVPN master and adds key features and bug fixes for a much more comfortable and reliable experience:

190 commits sounds like a pretty big divergence from the upstream OpenVPN 3. Are you planning to merge it to upstream or is it a hard fork?

[OpenSourcerer 1435]

There was a bit of a dispute between AirVPN and Arne Schwabe, the dev of OpenVPN for Android and maintainer of OpenVPN3, when AirVPN tried to upstream CHACHA20-POLY1305 when it wasn't implemented in v3 yet:Schwabe raised some points about code style for example to which AirVPN (or rather ProMIND) first adhered, but then a discussion emerged when they piled on. At the end, AirVPN closed the merge request and never again tried to upstream things to OpenVPN3. As Schwabe pointed out, it somewhat looks like AirVPN is proud to be multiple commits ahead of upstream, which is a weird stance to have in the FLOSS world.
I think it were even more awesome if AirVPN dev work would find official, worldwide recognition in OpenVPN3's commit history, but alas.

[Staff 9972]

21 hours ago, benfitita said:

@Staff In your Black Friday post you mention:
 
190 commits sounds like a pretty big divergence from the upstream OpenVPN 3. Are you planning to merge it to upstream or is it a hard fork?

Hello!

In 2019 we pushed a commit with a major new feature for OpenVPN3 library but it was refused by the maintainer (Arne Schwabe) for stylistic reasons. In the opinion of some of us it's their (OpenVPN3 maintainers) style to be ugly, but OK it's a matter of personal tastes so the text format and spacing was changed accordingly and a new commit was ready.

The new commit was again refused, this time because the source code included tags mentioning the source code author "ProMIND". We couldn't fully understand why the author should not be mentioned in the source code itself, moreover the tags were useful for another purpose, but fine, all tags were removed and a new commit was ready.

The new commit was again refused, this time because the identity of the source code author ProMIND was not certified. In general we do not disclose the identity of our employees and collaborators, much less force them to certify it with any third party, and leave to them the choice to disclose real identity and certify it. The request was strange and we asked for some clarification. It came out, even from other OpenVPN community members, that the requirement was related to a specific contractual agreement mentioned here: https://github.com/OpenVPN/openvpn3/blob/master/CLA.rst

At that point we did not like the situation: please note that a new problem was mentioned only after each new commit was proposed, while it would have been fair that all problems were mentioned at the same time, obviously, since the very beginning. And we did not like anymore to allow OpenVPN Inc. to re-license under any new license our or ProMIND's code, which is under GPLv3, according to the mentioned contributor's agreement, specifically part II, (e) clause:
"(e) I understand that OpenVPN Inc. may relicense this project, this contribution, and any modification to it under any license. [...]".

Therefore we did not waste additional time on the matter and we went on with our fork without further ado. At this very moment we have no merging/commit plans.

On a lighter tone, this thread dated 2019 is funny in the last part:
https://airvpn.org/forums/topic/43850-openvpn-3-development/
because one of the AirVPN community moderators defended OpenVPN style and Schwabe refusals, and wrote: "If you had contributed to the Linux kernel like that, Linus would tear you to tiny bits." Ironically, after that message was written, Linus Torvalds examined OpenVPN code and he was horrified.

Kind regards
 

[OpenSourcerer 1435]

15 hours ago, Staff said:

On a lighter tone, this thread dated 2019 is funny in the last part:
https://airvpn.org/forums/topic/43850-openvpn-3-development/
because one of the AirVPN community moderators defended OpenVPN style and Schwabe refusals, and wrote: "If you had contributed to the Linux kernel like that, Linus would tear you to tiny bits." Ironically, after that message was written, Linus Torvalds examined OpenVPN code and he was horrified

Mmh, your sass was duly noted.
If you pushed your code style onto the Linux kernel, it would've happened; what in god's name does that have to do with what Linus did or didn't say about the individual code style of a different project?

The much more funny thing is how overly sensitive you react to the slightest smell of criticism… it's irritating.

[Quallian 26]

On 11/25/2023 at 9:04 PM, Staff said:

Therefore we did not waste additional time on the matter and we went on with our fork without further ado. At this very moment we have no merging/commit plans.

I remember this soap opera in 2019. What I found disturbing and unprofessional was that each time you complied to some request a new problem was "found" by the maintainer, a wicked loop maximizing time waste. You were even too patient to re-push multiple times. You had implemented something new and important, and well tested, but the maintainer seemed focused to consider a problem a tab and a tag, before the identity certification requirement finally came out when all the "style" requests were satisfied and he went out of excuses to refuse.
Something similar happened to @NaDre (I hope the username is correct), he wrote extensively on it on this forum, when he pushed some patch to fix bug(s) of OpenVPN 2, but no merge followed and the bug stayed for years. Maybe Nadre can give us more details if he reads me.
I think this lack of professionalism could have discouraged external contributions to OpenVPN. No wonder we saw bugs haunting ovpn3 for years and years before AirVPN fork fixed them in few weeks. No wonder that important features of your forked library were implemented on the main branch with a delay of months or years (or never to date).

P.S.
The special authorization to let OpenVPN Inc. re-license arbitrarily anyone's contributions with any existing or future license may be disturbing and I can understand that you finally became reluctant to accept it.


 

[benfitita 39]

What a terrible story. Makes me wonder if some of these bugs or lack of features are intentionally kept in OpenVPN for some nefarious reasons.

I hope one day OpenVPN leadership changes and things can be back to normal. There's story of bcachefs that was recently successfully merge to mainline Linux kernel after around 10 years of development and lengthy discussions with multiple kernel devs.