Jordan, I have been following this work and I appreciate that very much. Your doc draws a good picture of the status of our codebase.
Personally I see much value in opening Zookeeper to non Java native clients. Reworking the internals (zkdatabase, server-to-server) as you state in your docs, is very dangerous and I am not sure it is worth to do in the short/mid term. The very trade-off we should accept will come when we decide how much efficiently non-jute client requests are to be processed. My mind is mostly over problems like zero-copy memory handling, saving resources on decode/encode. My other concern is about the concurrency model on clients. Zookeeper client API/contract relies heavily on a strict ordering of event delivery to the application. I feel we can implement this correctly but it won't be easy. To summarize I totally sponsor this work, your plan is reasonable, but I am not sure how much deep we can go inside the core of zk server. Starting with a gRPC endpoint is a good starting point Thank you for this hard work Enrico Il ven 17 gen 2020, 02:21 Jordan Zimmerman <randg...@apache.org> ha scritto: > Hello folks, > > I've been working on gRPC support for ZooKeeper. Please see the doc here > for latest details, links to branches, etc.: > > https://docs.google.com/document/d/1wP61nCDeLLNPdXGxHxi_MIbt7c2pCl0ttm9KMjyOisQ/edit# > > Also see: > > https://issues.apache.org/jira/browse/ZOOKEEPER-102?focusedCommentId=17017618&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-17017618 > > I'd appreciate some help, comments, etc. In particular, I need a ZooKeeper > committer to champion this. > > -Jordan >
