James, Maybe I'm missing something but I'm still not quite seeing how
protobuf helps us here.  It requires a "schema" of some sort and Gremlin
results are anything but a schema.  So you've told me that protobuf3
doesn't require that, but after a some limited review of the docs i don't
see where it behaves like that:

https://developers.google.com/protocol-buffers/docs/proto3

The best I could see was that it has an Any "data type" which can take
arbitrary bytes...is that what you meant?

> The Any message type lets you use messages as embedded types without
having their .proto definition

if so, doesn't that still leave us in a position where we still need to
serialize results to that blob?  no? I also saw that there is support for
Map:

https://developers.google.com/protocol-buffers/docs/proto3?hl=en#maps

but it looks like it must have homogeneous values, which might be a
limitation we could live with, though I am aware of production gremlin that
returns mixed value Map (same issue with List).

I think you're going to have to elaborate a bit more with some specifics.



On Thu, Dec 3, 2015 at 9:44 AM, Marko Rodriguez <[email protected]>
wrote:

> Hi James,
>
> I'm not so on the up-and-up when it comes to
> RPC/Protobuffs/MessagePack/WebSockets/etc. (are those even in the same
> category?! :).
>
> Can you say a little more on how this would help cross-language support?
> Are we saying something more than what we (Stephen) already do with
> WebSockets and the binary Gremlin protocol?
>
> http://tinkerpop.apache.org/docs/3.1.0-incubating/#_developing_a_driver
>
> Thanks,
> Marko.
>
> http://markorodriguez.com
>
> On Dec 2, 2015, at 3:24 PM, James Thornton <[email protected]>
> wrote:
>
> > Adding HTTP2 support via gRPC would give us solid, cross-language support
> > and allow us to auto-generate the client libraries http://www.grpc.io
> >
> > I had worked on a SPDY module, but HTTP2 has now become a standard since
> > usurped it.
> >
> > Thoughts?
> >
> > - James
> >
> >
> >
> > --
> > James Thornton, *http://electricspeed.com <http://electricspeed.com>*
>
>

Reply via email to