Proposal: release Savon v3 under a new name

[tjarratt]

Hey Savon users,

Under the issue for Savon v3 there's been some discussion around some of the breaking changes in the current unreleased version and whether it would be better to release it as a major version bump or as a separate gem. See: #473 (comment)

I've opened this issue for people to discuss the pros and cons of releasing Savon v3 under a new name. Some of these are obvious to me, so I've taken the liberty of listing these below:

Pro

• The current interface for Savon v3 is significantly different from v2
• A new repository would allow for separate issue trackers for v2, v3
• A clean start allows us to more easily drop support for older ruby versions, ruby implementations and some features with a high cost to maintain

Con

• resetting Savon's master branch back to v2 is a huge breaking change for any fork
• a new gem name fractures the community (eg: "Oh, which 'savon' are you using?")
• it is unclear what the future of Savon would be. Would it stay at v2 forever?
• backporting any fixes from v3 to v2 will only become more difficult with time

If you have a fork of Savon, or feel strongly either way, your thoughts and feedback would be greatly appreciated. Thanks!

[rubiii]

hey @tjarratt,

resetting master might cause problems, so thanks for creating an issue for this.

since version 3 is completely different from all previous versions and mainly focuses on providing a sane abstractions on top of wsdl and xsd documents, i would vote for moving version 3 to its own repository.

also, the issues associated with version 3 are quite different than the issues for version 2, because both versions are in a very different situation right now (stable/unstable) and also the architecture behind both versions is quite different, so you probably won’t just backport fixes to version 2.

by moving version 3 to its own repository and giving it a new name, we would allow for version 2 (savon) to evolve without people even having to know about a newer, different and unfinished version.

i’m really glad to see, that version 3 seems to already be used, but it’s not stable (meaning you should expect changes). maybe calling it version 3 makes it look more mature than it is.

[machielg]

I'd agree to anything that might spur the interest and development of v3. The legacy and unclarity coming from v2 is discouraging, my guess is people are more willing to try a beta of a new Gem than a beta of a gem that has a stable version.

[keltia]

Personally, I 'm fine either way although keeping the same repo would be easier indeed.

[tjarratt]

I spent some time today looking at usage of Savon on github and it seems like most people have it version locked in their Gemfile to ~> 2.x, so I'm feeling a lot less concerned about breaking people using 3.0. As @machielg said, the best thing we could do is help drive adoption and hacking on Savon 3, because driving more interest in v3 can only improve the community over time.

Will keep this open for some time longer to solicit feedback; I noticed some discussion on twitter today and if anyone is really truly going to be burdened by this change, I'd like to give them a chance to speak up. So glad that we've already had a conversation about this -- people really care about Savon :-D

[apenney]

I'm definitely in favor of renaming. I just attempted to switch to v3 only to find out how dramatically different it is. Especially considering some things that work in v2 seem unsupported in v3 (like rpc/encoded operations, which I guess all of the f5 stuff I'm trying to do is).

Seems like a rename and blaze forward without regard for older versions would be a good choice here.

[tjarratt]

Okay, it seems like there aren't any loud objections to reverting Savon master back to v2 and creating a new repo. Closing this for now.