WezM.net Menu

Goals and Directions for Rust in 2018

Published on Sun, 7 January 2018

This is my response to the call for community blog posts reflecting on Rust in 2017 and proposing goals and directions for 2018. See also readrust.net, where I’ve collected all the #Rust2018 posts.

2017 saw some great progress in the Rust space. The project had a clear roadmap for the year and followed it quite closely. It was a pleasure to see the outcomes of the 2017 survey systematically addressed in the roadmap.

Over the course of the year we saw some software outside of Firefox and developer tools (such as rustfmt, racer, rustup) gain wider use and make it into OS package archives. At the time of writing there are 17 ports dependent on rust in the FreeBSD ports tree, and 20 packages in the Arch Linux package repos including:

As I spoke about at the September Melbourne Rust Meetup, I think adoption of tools written in Rust outside the Rust community helps increase adoption of the language as whole. It means packaging systems are updated to support Rust, tools are exposed to more diverse environments, and people are more likely to raise bugs or contribute fixes and improvements. They may even learn some Rust in the process.

We also saw the initial release of Futures and Tokio — the initial answer to how Rust will support asynchronous I/O. I think these components will be essential in making Rust a viable option for people wanting to build network daemons and HTTP micro-services that can compete with the likes of Node.js and Go. I found the fact that Futures were architected in a way that upholds the abstraction without overhead (zero-cost abstractions) goal of Rust to be particularly impressive.

2018

With most of 2018 ahead there are four areas that I would like to see Rust improve on this year:

  1. Become a better option for building network daemons and HTTP micro-services.
  2. Continue to improve the discoverability and approachability of crates and Rust’s web presence in general.
  3. Gain wider, more diverse tier-1 platform support (especially on servers).
  4. Start delivering on the prospect of safer system components, with fewer security holes.

Network Services

There’s a lot of interest these days in HTTP micro-services, often running in containers on clusters. This is an area that Go does quite well at. Its minimal runtime, language level async support, core http library, and easy cross compilation make it well suited to this task. Node.js too can work well for these types of services. On the face of it Rust should be as good or better than Go and Node at these types of services with its even smaller runtime footprint, sophisticated type system and extra guarantees. However, until Tokio matures it’s not really in the running.

I would like to see work continue on stabilising async/await, stabilisation of Hyper and support for HTTP/2. So that when a decision is being made about which technology to use for these types of services Rust is one of the contenders.

Improve Rust’s Web Presence

This is in part a continuation of the 2017 goal, Rust should provide easy access to high quality crates and the, Improve the Approachability of the Design of rust-lang.org and/or crates.io proposal in the Increasing Rust’s Reach program.

Having a consistent, approachable, discoverable, and well designed web presence makes it easier for visitors to find what they’re looking for and adds signals of credibility, attention to detail, and production readiness to the project. It would be wonderful to see the proposal above picked up and completed.

The large amount of Rust code that already exists is not particularly visible unless you know to look on crates.io or GitHub. Crates and Rust libraries should be more discoverable through traditional search engines. Aside from the static documentation, no pages on crates.io show up as search results on DuckDuckGo. They do show up in Google but the titles and descriptions shown often aren’t super useful.

When posting links to crates on Twitter or Slack, there is no rich preview, which would help potential visitors know more about the link and what they will find there.

Screenshot showing how a crate looks when tweeted.
How a crate looks when tweeted.

Screenshot showing how an npm package looks when tweeted.
How an npm package looks when tweeted.

The Rust Cookbook is an excellent resource, especially for those just starting out. It should graduate from the nursery and be made more discoverable, perhaps by integrating it with the categories on crates.io.

Platform Support

Currently there are three OSes with tier 1 support: Linux, macOS, and Windows. These are the three most popular OSes and it’s great that they’re all supported. However, it would be great to see more OSes gain tier 1 support. Platform diversity makes Rust a viable option for more projects and can also help find bugs.

Personally I would like to see FreeBSD promoted to tier 1 support. This would be difficult at first as CI infrastructure would need to be built, contributors would need to learn to address issues that would break the build on FreeBSD, etc. but in the end it would be easier to add more platforms in the future and the ecosystem would be more robust for it.

System Components and Increased Safety

One of Rust’s strengths is memory safety. Jokes about, “Rewrite it Rust”, and the Rust Evangelism Strike Force aside there has been a lot of talk about the possibility for Rust to prevent some common causes of vulnerabilities in C and C++ code. It would be nice to see some of this talk turn into action. I’m not talking about rewiting Linux or cURL in Rust but start with replacing some high value parts of existing C libraries like what Federico Mena-Quintero has done with librsvg.

Maybe these start as forks/experimental branches that can be used as drop in substitutes for the original so that adventurous users on bleeding edge systems like Arch, FreeBSD, or Gentoo could test them out.

While this is perhaps less of a goal for the Rust project and more of one for the Rust community there are parts of it that relate to the project. For example building these hybrid libraries would surely exercise and provide feedback for the ongoing task to improve the ability for Rust projects to integrate with existing build systems.

Conclusion

So they are my hopes and dreams for Rust in 2018. I feel that in some ways I’m just sitting back telling other people about all the hard work they should do. On the other hand I feel like I am contributing as well by writing and publishing Rust code, doing talks, and providing bug reports and pull requests.

Here’s to another successful year of Rust ahead!

For more great #Rust2018 posts check out readrust.net.

Comment icon Stay in touch!

Follow me on Twitter, or Mastodon, subscribe to the feed, or send me an email.

Short URL: http://j.mp/2CCHJ48