gaussianrecurrence commented on pull request #715: URL: https://github.com/apache/geode-native/pull/715#issuecomment-785976885
> I wonder if anyone of this is actually necessary. I think this idea of clearing the PDX registry is a hold over from the old global world. Under that old model if you "closed" the cache it was never really closed, especially from the .NET layer, thus it never really cleared the the PDX registry, which was also global. Now the PDX registry is owned by Cache and Cache can actually be closed. If the Cache is closed the registry is destroyed. I really think we should just eliminate all this code. After a detailed analysis of the problem we were experimenting, quite the opposite, this is much necessary. In fact there is the same behaviour in the Java client. This behaviour guarantees that if the cluster is restarted and the PdxTypeRegistry in the distributed system changes there is no de-sync between the client and the server. It's quite important that there are not PdxType in the client which does not exist in the server. Otherwise it could happen that some PDX entries with this PdxType are written in the server, corrupting the data in the region. TBH, I'd rather prefer that written entries which are PdxInstances are verified on the server-side, but this would probably increase put time. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
