[openstack-dev] [wsme] [ironic] [ceilometer] [magnum] [kite] [tuskar] WSME unmaintained ?

Chris Dent chdent at redhat.com
Thu Apr 16 16:45:03 UTC 2015


> On 04/16/2015 07:41 AM, Lucas Alvares Gomes wrote:
>> We have a couple of Openstack projects that uses WSME for their REST
>> APIs[1], but WSME project looks abandoned. The review stats are not good,
>> for the last 40 days the project didn't have a single review from a core
>> reviewer[2], the bugs are not being triaged nor fixed [3], I have been
>> trying to get some core people to look at a review for the last ~2 weeks by
>> asking for reviews in the #wsme channel without success. Plus, if you look
>> at the merged patches queue you will see that some people approve their own
>> patches*.

I've struggled with these issues too, largely because if you throw
anything other than a request that WSME expects at WSME, at worst
it breaks down and cries, at best it does something just plain weird.
I've tried to make some changes that will fix it, but a fix over here
breaks a thing over there (e.g. [1]).

>> * Should projects relying on WSME start thinking about migrating their APIs
>> to another technology?

I think this may be the way to go. The intent of WSME is admirable
(the multiprotocol stack) but the execution leads to unwarranted
complexity.

I think a framework more specifically dedicated to JSON APIs, or
even just being "webby" and correct would be better. Something
_much_ simpler and _much_ more aligned with WSGI.

One thing I would _love_ is for us to get away from object dispatch
and use explicit routing. Because it's, uh, explicit. But that's a
personal preference.

>> * Can we somehow get the core team to start paying more attention to the
>> project? Or can we elect some people willing to do some review to the core
>> team ? If so, there's anyone out there that wants help with it?

If we do choose to stick with WSME (which might be a good idea given
the number of projects using it) I think it will require quite a lot
of attention to make it work well[2] and thus would need some active
contributors.

>> * Forking the project an option?

In the context we're in wouldn't a friendly[3] takeover of the infra
config do the same job?

On Thu, 16 Apr 2015, Monty Taylor wrote:
> Or - as you bring up - perhaps our use of this has wound up having been
> a mistake and it's time to cut the cords. The original intent was to get
> all of openstack using the same framework, but this has not come to pass. :(

Was there any hope that would be fulfilled by having everyone on the
same framework other than "it's more tidy"?

[1] https://review.openstack.org/#/c/172520/

[2] Where well means actually following the rules of HTTP, not just
handling requests that are specially structured to be okay with
WSME.

[3] Or hostile if that were really needed.
-- 
Chris Dent tw:@anticdent freenode:cdent
https://tank.peermore.com/tanks/cdent



More information about the OpenStack-dev mailing list