[swift-evolution] Large Proposal: Non-Standard Libraries

Tue Nov 7 15:58:39 CST 2017

Hi Dave,

Thanks for bringing up this topic.  This has been kicked around a little, and we’re still exploring different models on how to extend Swift.

The server APIs work group is one operational model for the community to build out a new set of core libraries.  That work group was formed out of the observation that there were different implementations of the same thing being created by different Swift on server efforts, and that those different efforts should pool those resources together and standardize on some fundamentals.

That analogy could work here.  However, it also has possible major downsides.  It can be heavyweight, and also artificially raise the importance of certain people’s library efforts over others by creating a formal work group whose efforts are expected to eventually be incorporated into the core Swift distribution.  It would also force a discussion up front of what even makes sense to incorporate into the core Swift distribution, which might be artificially constraining the exploration of the set of libraries to implement.

I need to think about your idea more, but my first reaction is that my preferred route is that the community builds these libraries, ideally using Swift Packages, and through trial and use we evaluate those libraries and then decide if they should be incorporated as part of the core distribution.  There’s many factors involved in deciding if they can be incorporated into the core distribution, but all of those could be informed by actually building libraries and see them getting used.

We could also figure out a way to possibly highlight these efforts to the Swift community, maybe on swift-evolution or other means — but all with the expectation that these libraries are not *necessarily* going to be standardized as part of the core swift distribution.  I don’t think that’s all that bad either; not every library will make sense to incorporate into the core Swift distribution (e.g., they are highly domain specific) but still supporting their development would be beneficial to the community.

Note that any change going into the Swift core distribution inevitably involves swift-evolution and the core team.  Changes going into the core Swift distribution must meet a very high bar as far as implementation and design, the confidence we have into committing to specific APIs (especially since we’ll have ABI stability in Swift 5), and other factors.  Thus this will not really alleviate much burden on the Core team or on the community — one of the core goals of your proposal.  Further, incorporating all those concerns up front when building libraries might be counterproductive.

FWIW, Ben Cohen and I have been talking about possibly using Swift packages as a way to seed out experimental ideas for extensions to the Standard Library.  This would allow ideas to be trialed by real usage (a complaint I’ve seen about some changes we’ve made to Swift in the past).  Users could build things on top of those libraries, knowing they are available as packages, and if an API “graduates” to being part of the Standard Library the user can then depend upon it being available there.  If it never graduates, however, the package remains around.

One thing that resonates me in what you propose is about having a “well-organized effort” whose aim is to make these libraries feel cohesive and implemented well.  However, given that many of these naturally fall somewhere in the spectrum of responsibilities within the Standard Library and Foundation, I think it is self-misleading to think that the Core Team or others would not be involved in these efforts if the intention is to possibly incorporate these one day into the core Swift distribution.  There also may be other ways to achieve that level of cohesion in API design, such as through community discussion (possibly via the Swift.org <http://swift.org/> mailing lists/forums).  This discussion would not necessarily be the same as the path to “ratification” of a library into core Swift, but a step that could benefit many library authors (including considering ideas one day incorporated into the core swift).  With such a mechanism many library authors could benefit even if they do not intend to have the library as part of the core distribution.

My apologies that these are all hand-wavy ideas, but I’m trying to paint a possible alternative way to achieve your goal of more library evolution for Swift without having such a formal work group that may be too heavy weight or too narrowly focused.

Ted

> On Nov 7, 2017, at 9:54 AM, Dave DeLong via swift-evolution <swift-evolution at swift.org> wrote:
> 
> Hi Swift-Evolution,
> 
> The Standard Library's goal is to be small and targeted. However, many aspects of Apple-provided frameworks need or offer opportunities for improvement or wholesale replacement. These enhancements lie beyond the scope of the Standard Library.
> 
> To address this, we'd like to propose the idea of a "Non-Standard Library"; this would be a library that ships with a regular installation of Swift, but is not imported into .swift files by the compiler, unless explicitly requested by the developer.
> 
> We are proposing a well-organized effort to parallel the Standard Library without putting additional implementation responsibilities onto the core team. This effort would mitigate what we see as platform-independent requirements that provide native Swift implementations that aren't burdened by Apple history.
> 
> Mission Statement
> 
> We propose to create an open-sourced "Non-Standard Library" effort that coexists with, coordinates with, and is blessed by the open source Swift development community. The "Non-Standard Library" will offer a well-curated, high-value collection of routines, types, and documentation to support Swift development on all platforms.
> 
> Goals
> 
> The main goals of this effort would be:
> Alleviate pressure on the Core Team while providing the developer community with functionality that normally falls under Apple's internal development umbrella.
> Deliver authoritative, reliable, and regularly updated libraries avoiding issues faced by third-party dependencies.
> Provide oversight, organization, and full community involvement to ensure its components are worthy, maintained, and passing a bar of need and excellence.
> Suggested Libraries
> 
> There are several areas we think that are ripe for improvement and implementation as a non-standard library, such as (but not limited to):
> A BigNum library
> A full-featured Random library
> A simplified date/time library
> A library for manipulating paths (that is not based on URL or String)
> An expanded Swift-native Geometry library and so forth.
> The scope and extent of the sublibraries would be proposed and debated on a parallel Non-Standard Library Evolution development list.
> 
> Coordination
> 
> This effort would be fully coordinated with the Swift development team, respecting the team's existing commitments and responsibilities. We would request an Apple body to act as an official coordinator, enabling both oversight of this effort and to expose Apple-sourced suggestions to the Non-Standard community for action.
> 
> Next Steps
> 
> To proceed, we need a general community consensus that this effort is worth the significant overhead it would involve.
> 
> We must then:
> Select a project lead, who appoints technical leaders from the community.
> Recruit a core team, a small group responsible for strategic direction, pulled from experienced members well versed in contributing to Swift-dev.
> Establish a Non-Standard Swift Evolution process, based on the one that is currently followed by Swift Evolution. Following SE practices will guarantee a consistent and excellent language experience for those developers including the Non-Standard library into their projects.
> Build a Non-Standard Swift Evolution repository home.
> Adopt a code license, based on Swift's current licensing.
> Define the community working group rules and code of conduct.
> Establish a mailing list and/or forum.
> Begin the selection of target sublibraries and populating them by recruiting committers and contributors.
> Alternative Names
> 
> Suggested names for this effort include the following.
> Extended-Standard Library
> Community-Standard Library
> Non-Standard Library
> We are not married to any of these names and welcome alternative suggestions.
> 
> Measures of Success
> 
> A successful Non-Standard Library will provide a stable and incrementally grown ecology of Swift frameworks that developers can depend upon with a minimum of turbulence and regret cycles. For that reason, the most successful content will be core functionality, with significant field testing prior to adoption. We recommend that NSSE follow a model of provisionary adoption and refinement before deployment to ensure any adopting code base will not be affected by later module changes.
> 
> Thanks,
> 
> Dave DeLong and Erica Sadun
> _______________________________________________
> swift-evolution mailing list
> swift-evolution at swift.org
> https://lists.swift.org/mailman/listinfo/swift-evolution

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.swift.org/pipermail/swift-evolution/attachments/20171107/4f4f438a/attachment.html>