NOnostrademons·May 17, 2016
You probably would add the most value by improving the maturity level of existing libraries. Some that I've run across that exist but don't really give me much confidence in using them in production include:
https://github.com/sbeckeriv/warc
http://techern.github.io/Varint-rs/varint/index.html
https://github.com/hyperium/hyper
https://github.com/rusoto/rusoto
https://github.com/servo/html5ever (largely on the performance front; on a recent test of GZip + WARC reading + HTML parsing, Kotlin (on the JVM) + WebArchive Commons + JSoup beat Rust + warc_nom_parser + html5ever by about 25%, which does not fit with Rust's reputation as a fast systems programming language)
And libraries that I wish existed but which don't really have good options:
Web crawling. Hyper is listed above as an HTTP library, but there's nothing like Apache Nutch which
Read the rest
respects hostloads and robots.txt and works as a turn-key solution.
RSS parsing. RSS serialization works, and people have written toy RSS readers, but nobody's written a robust, self-contained RSS parsing library.
Machine learning. It was a shame when Autumn.io gave up Leaf.
Big data. Create a Hadoop/Spark alternative in Rust and the world will love you, at least if you can beat Hadoop's reliability, performance, and ease-of-use.