On 22/06/17 18:50, Michael Vorburger wrote: > Hello, > > Naive Q: Do we already have gRPC/Protocol Buffers (PB) support for ODL > RPCs? Say if you were writing something in say Go and wanted to call > some ODL RPC, and sure could start with a HTTP REST call to the RPC's > RESTCONF, but wondered whether there was a faster way with generated > strongly typed client library for the future... interesting/useful? Stupid?
The trouble with protobufs is that they are inherently tied to a fixed set of data models (something like a chunk of SchemaContext with augments resolved), hence they cannot be readily reused with XML Infoset. > If one were to implement something like this, it would best be... > presumably, a new generator, a bit like a binding, which would generate > a *.proto for a *.yang (no idea how easy/hard this is?), and then use > the PB infra to gen. the clients, and then gen. an impl of the PB server > "stubs" mapping the request/response from PB to a Java Binding or > (probably better) Normalized Node - does this idea make sense, in principle? This is your basic schema-informed coding. Protobufs make sense if: - you have a fixed set of immutable models (not counting in features and deviations), or - solve the extensibility problem somehow, and - can deal with upgrades (which means new, probably incompatible .proto files) Other than that it does make sense and there is EXI, which (obviously) handles all of XML Infoset -- but that does need further work from upstreams and inside ODL: - either get Exificient to finally release 0.9.7 or improve OpenEXI to have a saner grammar integration surface - finish up https://git.opendaylight.org/gerrit/#/c/35136/ Bye, Robert
signature.asc
Description: OpenPGP digital signature
_______________________________________________ controller-dev mailing list controller-dev@lists.opendaylight.org https://lists.opendaylight.org/mailman/listinfo/controller-dev
