An update here is that we are beginning development of a specific proposal, and will keep the ticket and this mailing list thread up to date with our status.
On Thu, Sep 14, 2023 at 7:56 AM Raphael Taylor-Davies <r.taylordav...@googlemail.com.invalid> wrote: > Hi, > > Thank you for starting this discussion. I think the decision to use gRPC > and by extension HTTP certainly would encourage a design that explicitly > doesn't rely on server-side state. Not only is it now uncommon for > backend servers to have a unique globally-routable identity, but as > these are request-oriented protocols, as opposed to connection-oriented > protocols, maintaining session state server-side becomes very > complicated as there is no unambiguous end-of-session signal. > > I would very much encourage following an approach similar to web > cookies, where state is instead managed by the clients and sent with > each request. This sort of maps to the ticket notion already present in > many of the APIs, but could perhaps be formalized. > > Kind Regards, > > Raphael Taylor-Davies > > On 14/09/2023 11:52, Andrew Lamb wrote: > > Hello, > > > > As FlightSQL gets more widely adopted across the ecosystem, we hit some > > issues trying to implement bind parameters in our stateless service. I > > filed a ticket [1] that describes the issue as well as a potential > > solution. > > > > Please share your thoughts on the ticket > > > > Andrew > > > > [1] https://github.com/apache/arrow/issues/37720 > > >
