grpc-web VS sqlx

A proxy is required to communicate from web clients to a server running gRPC, and there are only two choices for this proxy: a. The Improbable gRPC-Web client or b. The Google gRPC-Web client

I'll have to try the Improbable method (https://github.com/improbable-eng/grpc-web) - as it definitely includes a websocket transport, hopefully this multiplexes requests on that transport and would hence solve the problem. I'll post again once I know...

I think you should check out https://github.com/improbable-eng/grpc-web/tree/master/go/grpcwebproxy

What I like with gRPC is that it allows both client and server side streaming. Twirp seems to not have this feature at all. In contrast the https://github.com/improbable-eng/grpc-web even supports server and client streaming while wrapping gRPC in a HTTP1.1 Web connection.

I highly recommend also to look into https://github.com/improbable-eng/grpc-web in contrast to the 'official' gRPC-web from google the improbable-eng Implementation can do both server side and client side gRPC streaming (when used with their JavaScript client library) and all this on an http1.1/websocket basis...

We use https://github.com/improbable-eng/grpc-web for the server part with https://github.com/protobufjs/protobuf.js on the frontend.

I'm debating whether or not to rely on grpc for all client to API transport using improbable-eng's grpc-web project. The hesitation I'm running into is concern over unforeseen pitfalls. The fact that it isn't widely used is the main cause for my apprehension.

If you use https://github.com/improbable-eng/grpc-web you do not need envoy, you can wrap your golang grpc connection. I then mux the grpc and grpc-web so that I can use the same endpoint to serve http1 and http2 clients. This makes it easy to host on the cloud, eg. google cloud run.

I'm setting up a server portion of a website which uses (g)RPC. My dev setup is a docker-compose.yml that starts up my server on 50051 & another service that runs grpcwebproxy (exposes 8080 and proxies to my server @ 50051).

For PostgreSQL, the most relevent part of the code is here. With this in mind I changed some things around to rely on schemas instead of databases and even simplified some parts of the implementation as this was always meant to be for internal use only..

What exactly are you missing? I haven't really written "boring corporate backend stuff" in a few years but something like sqlx provides everything I've ever needed there.

Badges are the little rectangles you typically see at the top of a crate's README: https://github.com/launchbadge/sqlx/blob/main/README.md

SQLX has entered the chat [1].

[1] https://github.com/launchbadge/sqlx

Database : SqLite (using sqlx).

The current (well, removed) MSSQL sqlx driver also doesn't support encrypted connections (https://github.com/launchbadge/sqlx/issues/1552), so it might work locally on an unnamed instance but not over the network until they add in the newer driver (which will be closed source but for an OSS project you should be able to request a free license I think)

I'm considering making my own wrapper over lower-level database drivers. There are a lot of features in sqlx that I don't need, and the latest version seems to have removed useful data structures that SQLPage is using. It also removed support for SQL Server.

add Ed25519 to AuthPlugin enum on https://github.com/launchbadge/sqlx/blob/061fdcabd72896d9bc3abb4ea4af6712a04bc0a8/sqlx-core/src/mysql/connection/auth.rs and implement them using these crates: