Re:Fractalide: Simple Rust Microservices
by setori88 on Friday December 09, 2016 @11:17AM
(#13139581)
An author here: we're working on the nixcrates project to remove cargo and allow for closer integration between nix and crates.io, so far you can check out the project at https://github.com/nixcloud/nixcrates (joachim and paul are super people, hit them up if you're interested in contracting them) the net/http stuff is currently incompatible with the most recent changeset, I'll get working on it tomorrow morning HKT to make it work, I have to refactor it to suit the new stabilized API. Anyway, this attention was a bit earlier than expected, still have a couple of weeks to go before stabilizing. I hope you guys can kick the tires, tell us what you don't understand and please comment on the API because we want something sensible. We're using the late Pieter Hintjens' C4 so please hop on board and collaborate!
Re:Fractalide: Simple Rust Microservices
by setori88 on Friday December 09, 2016 @07:58PM
(#13143301)
There is too much cognitive dissonance between nix and cargo. It might be better to get nix, a package manager, to do the job of cargo, a package manager.
Re:Fractalide: Simple Rust Microservices
by akubera on Friday December 09, 2016 @10:55AM
(#13139409)
Most documentation is in the form of README files within the subdirectories; here is the code auto-docs: https://docs.rs/fractalide
Looks like knowledge of the nix distribution system is required to use. I don't know how hard it would be, but a neat auxiliary project would be a cargo plugin that supports fetching packages though nix.
There is an example TODO app in the https://github.com/fractalide root dir (split between fractal_app_todo_{,model,controller}).
Upon further inspection, this seems to be "simple nix microservices" that happen to be implemented in rust. One of the goals is to "remove cargo", and they use `nix-build` over `cargo build`. Also, a lot of implementation is in their 'flowscript' language within a lot of *.nix files [1] [2].
Interesting project nonetheless. Good luck to them!
Re:Fractalide: Simple Rust Microservices
by eclark on Friday December 09, 2016 @12:57PM
(#13140372)
Same. Right now rust feels pretty isolated on the network.
There's no great RPC support. No thrift, grpc, etc.
The http libraries are good for low level, unsecured, stateless services. However they lack lots of more complex features.
Re:Fractalide: Simple Rust Microservices
by setori88 on Friday December 09, 2016 @07:04PM
(#13143006)
a) yes it will, we'll have to do some design on the `fvm` to make it hotswap. It won't be ready for this stabilization release. The functionality is already there it just needs
b) I'm not really focusing on other market occupants. We need this system to solve some pretty hairy problems 100TB problems.
c) 'feature-credible, testable release' primarily this is dependent on `nixcrates`, documentation and maybe getting a few more services examples. I'd say a couple of weeks.
Looks like knowledge of the nix distribution system is required to use. I don't know how hard it would be, but a neat auxiliary project would be a cargo plugin that supports fetching packages though nix.
The "net/http" fractal definition [0] references the repository https://github.com/fractalide/fractal_net_http, which is where the rust microservice code actualy lives.
There is an example TODO app in the https://github.com/fractalide root dir (split between fractal_app_todo_{,model,controller}).
Upon further inspection, this seems to be "simple nix microservices" that happen to be implemented in rust. One of the goals is to "remove cargo", and they use `nix-build` over `cargo build`. Also, a lot of implementation is in their 'flowscript' language within a lot of *.nix files [1] [2].
Interesting project nonetheless. Good luck to them!
[0]: https://gitlab.com/fractalide/fractalide/blob/master/fractal...
[1]: https://github.com/fractalide/fractal_app_todo/blob/master/c...
[2]: https://github.com/fractalide/fractal_app_todo/blob/master/c...
a) will Fractalide support dynamic dataflow graph change at runtime?
b) For reasons of clarifying your objectives further, could you elaborate on your target market's current occupants? Is this Kafka / Flink for Rust?
c) What is your estimated timeframe for a feature-credible, testable release, even if beta?
b) I'm not really focusing on other market occupants. We need this system to solve some pretty hairy problems 100TB problems.
c) 'feature-credible, testable release' primarily this is dependent on `nixcrates`, documentation and maybe getting a few more services examples. I'd say a couple of weeks.