Log file Viewer

[20:03:55] <hazmat> J1m, i'm starting to think the zc.zk behavior of re-establishing ephemerals shouldn't apply at all to sequence ephemerals

[20:04:12] <hazmat> it breaks most of the coordination primitives recipes

[20:13:10] <J1m> I think you're right.

[20:13:16] <J1m> hazmat, ^

[20:17:08] <J1m> Note that that behavior is easy to bypass now by calling the low-level api with the .handle attribute.

[20:17:44] <J1m> I hadn't paid much attention to the sequence part of the API.

[20:18:11] <J1m> That would be easy to fix and would provide a good reason to add sequence support to the testing mock.

[20:19:29] <hazmat> J1m, are all the tests mock based?

[20:19:57] <hazmat> looks that way..

[20:20:47] <hazmat> the question is even removing the sequence ephemerals from the session tracking.. how does the higher level primitive get notified of the session expiration and its loss of the lock

[20:21:02] <hazmat> er.. lock/eph sequence node

[20:25:48] <hazmat> ah.. ic the env var lets it test against a real one

[20:34:52] <J1m> right

[20:35:47] <J1m> Not sure if there's still an unanswered question above. :)

[20:36:35] <J1m> btw, Ben Bangert was interested in separating the service-management bits of zc.zk from the basic bits.

[20:37:11] <J1m> I was thinking of splitting it into some packages in the zk namespace, ala zk.base, zk.service, zc.locking...

[20:37:41] <J1m> hazmat, ^

[20:41:30] <hazmat> J1m, that sounds nice.. i've been talking to him as well

[20:42:04] <hazmat> txzk has all the basic recipes, but twisted seems to big a divide for most

[20:43:00] <hazmat> J1m, i've sort of fallen away from fine super grained packages, a single package for zk.recipes would suffice for most i think (queue, locking, etc)

[20:43:24] <hazmat> zk.service is pretty app specific though so living its own pkg makes sense.

[20:43:50] <J1m> yup

[20:44:20] <J1m> BTW, I'm using zc.zk with a gevent based app. :)

[20:44:31] <J1m> but you knew that, I guess

[20:46:54] <hazmat> J1m, indeed, i like how the code naturally flows for both usages (greenlet or standard sync)

[20:47:18] <hazmat> J1m, does that mean use the async apis for gevent apps?

[20:47:29] <hazmat> ^you

[20:47:37] <J1m> no

[20:48:05] <J1m> I could, but I don't really need to.

[20:53:16] <hazmat> J1m, ic.. that reminds me if you do i recently had to patch a bug for the python bindings and async usage.. they leaked.. https://issues.apache.org/jira/browse/ZOOKEEPER-1431

[20:53:30] <hazmat> it will be in the zk packages in precise

[21:00:26] <J1m> good to know.

[23:14:52] <aclark> jaraco: eggmonster!

[23:15:14] <aclark> !motivate jaraco

[23:15:14] <pmxbot> you're doing good work, jaraco!