Akshay Shah
Akshay Shah
> The aliased types should be generated only for the base package, so if any request/response type isn't in the same proto package then the alias doesn't exist... This seems...
@marekbuild That's nice and avoids conflicts, but it generates code into the base types - which we don't want to do, because it's messy and incompatible with BSR remote packages.
We've let this sit for quite a while. I think we can move forward with generating type aliases in the Connect packages. Regarding Peter's concern with name collisions, I think...
Do we need the service name prefix? Just this should be fine: ```go Ping(context.Context, *PingRequest) (*PingResponse, error) Fail(context.Context, *FailRequest) (*FailResponse, error) Sum(context.Context, *SumStream) (*SumResponse, error) CountUp(context.Context, *CountUpRequest, *CountUpStream) error CumSum(context.Context,...
I've opened #560 - it's a little different from @emcfarlane's approach. In general, it's more conservative and only generates aliases for the well-behaved subset of messages where we can safely...
@meling Generating aliases for the stream types isn't infeasible, it's just more complex - there are more naming decisions and potential conflicts to worry about, so there's more to debate....
#560 doesn't have majority approval either 🤣 In the discussion, @mattrobenolt made a good suggestion to consider using custom options: https://github.com/connectrpc/connect-go/pull/560#issuecomment-1673580150 Pursuing that approach is probably the next step here.
On the client side, users can already use `net/http/httptrace`. On the server side, users can employ one of the _many_ authn/authz packages that wrap `http.Handler` - and if they like,...
Sorry for the confusion, @ImSingee - this issue hasn't reached enough consensus for PRs. > Though I do wonder what the value of Peer is in its current form. It's...
As an update, I've plowed some more time into my standalone authentication package this weekend. It's now [`go.akshayshah.org/connectauth`](https://pkg.go.dev/go.akshayshah.org/connectauth), and it looks a bit more reasonable to me. In particular, the...