Julian Hyde wrote:
On Nov 8, 2016, at 4:15 PM, F21<[email protected]>  wrote:

On 9/11/2016 11:06 AM, Julian Hyde wrote:
A couple more things:
* Integrate the driver and its documentation into Avatica’s web site. I agree 
with putting source code in a separate repo, but all drivers should be on the 
same web site.
This can be easily done for the readme and usage docs. For the API docs, the 
godoc tool can generate html files from the source code, however for opensource 
projects, we tend to defer that to godoc.org. For example, if the project is 
hosted at github.com/Boostport/avatica, then going to 
godoc.org/github.com/Boostport/avatica would automatically pull in the source 
and generate the documents. Would it be acceptable to defer the API docs to 
godoc.org?

Yes, entirely acceptable.

+1 duh. Documentation is rather important :)

We must think of the consumers of our software, and deliver it in the way most 
convenient for them. Which means adhering to the norms of their community.

* The driver should know which ranges of protocol versions it can handle, and 
give a useful error if it encounters a protocol version outside that range.
There doesn't appear to be any versioning for the Protobuf protocol (I am 
assuming that with Josh's work, the goal is to make using old fields compatible 
and deprecate them). Can you expand on this one?

I don’t know enough about Protobuf to answer that one. But if Protobuf doesn’t 
know its version, at least an Avatica server does. the Go client is connecting 
to an Avatica server whose version is too or too recent, a user can reasonably 
expect a descriptive and useful error message, and I assume that it is 
technically possible to make that happen. If I’m wrong, can you correct me, 
Josh.

Julian


Right now, we haven't had to build such a thing. Wire protocol versioning is likely going to be a pre-req when we go to Avatica-2.0 (or whichever version we decide to start removing things that we've deprecated). Thus far, clients don't need to do anything different WRT the version of the server being communicated with; that's the "power" of protobuf.

Does that help clarify things?

Reply via email to