Open Stack

On Sun, Jun 11, 2017 at 10:47 AM, Robyn Bergeron <rbergero at redhat.com>
wrote:

> 2: The Github API is making some changes in moving from V3 to V4. Not
> clear to me if (a) they will live in coexistence or not, either forever-ish
> or otherwise, (b) if otherwise, when v3 will be eol'd, but might be worth
> considering if v4 helps the particular cases discussed in this thread in
> any way (i dont count "annoying jlk endlessly since he just finished all
> that github work," *hugs* to jlk tho.)
>
> If none of that is useful... sorry for the noise :)
>

We talked about this some on IRC. What Github is doing is (eventually)
retiring their REST based API that is v3, and replacing it with a GraphQL
based API that is v4. They build their own UI with GraphQL and everything
in the REST API is bolted on after they do it for themselves. They want to
end that, and instead let everybody play in the GraphQL pool and get access
to new features as soon as the UI is using them.

GraphQL is an interesting beast, from what I've learned so far, and it may
have some advantages in using it as a framework for servicing any Zuul API
like capabilities (including the status links). It does allow clients to
decide just how much data they want in a single grab, either a sip of the
status, or the entire firehose. Being able to do it in one API call reduces
resource consumption all around.

That said, I don't want to interject this particular bike into the shed
that's being built. I don't know enough about the pros/cons of GraphQL, let
alone the Python frameworks available to make use of, nor do I know enough
about the problem space that Monty is trying to solve.


- jlk
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-infra/attachments/20170611/e2138eb0/attachment.html>