Being protocol compatible is one thing, what you do in implementation is another. You may choose to a) Give a warning and do a best effort and b) Fail with appropriate error. The fact that protocol is compatible doesn't mean that the implementation must provide the full functionality as well.
On Fri, Jun 9, 2017 at 9:28 AM, Sergio Pena <[email protected]> wrote: > > > > On June 9, 2017, 6:09 a.m., Alexander Kolbasov wrote: > > > > > > > Sergio Pena wrote: > > Sasha, I think that making HDFS sync call compatible might cause an > inconsistent permission state to those callers who want to use the old API. > With the new behavior of HMSFollower to detect new full snapshots and > cleaning all delta changes to start fresh, now these old callers will > receive gaps of notifications. Do we really want to make this thrift > compatible with those old callers? And, if we want to push this new change > to sentry 2.0, then we are free to break compatibility here, don't we? > > Btw, if we do allow this compatibility on Sentry 2.0, then we will need to > maintain it in every Sentry 2.x version. > > > - Sergio > > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/59895/#review177445 > ----------------------------------------------------------- > > > On June 8, 2017, 5:42 p.m., Sergio Pena wrote: > > > > ----------------------------------------------------------- > > This is an automatically generated e-mail. To reply, visit: > > https://reviews.apache.org/r/59895/ > > ----------------------------------------------------------- > > > > (Updated June 8, 2017, 5:42 p.m.) > > > > > > Review request for sentry, Alexander Kolbasov, kalyan kumar kalvagadda, > and Vamsee Yarlagadda. > > > > > > Bugs: SENTRY-1782 > > https://issues.apache.org/jira/browse/SENTRY-1782 > > > > > > Repository: sentry > > > > > > Description > > ------- > > > > This adds the HMS image ID to the hdfs/sentry thrift definition. The > value will be used as part of the hdfs/sentry updates requests in order to > detect if a full paths image should be returned or not. > > > > > > Diffs > > ----- > > > > sentry-hdfs/sentry-hdfs-common/src/gen/thrift/gen- > javabean/org/apache/sentry/hdfs/service/thrift/SentryHDFSService.java > 29178d90b744b4b82b92ea24b1dc4f06dbebe36a > > sentry-hdfs/sentry-hdfs-common/src/gen/thrift/gen- > javabean/org/apache/sentry/hdfs/service/thrift/TPathsUpdate.java > f3014be9fec01bb32fa171d27bed9ac23bac4717 > > > > sentry-hdfs/sentry-hdfs-common/src/main/java/org/apache/sentry/hdfs/PathsUpdate.java > 6b31f7a4bbff66f6640813774ef79958990d7a57 > > > > sentry-hdfs/sentry-hdfs-common/src/main/resources/sentry_hdfs_service.thrift > d01841b3bc4a8b38f44e08b45449e11352f6c725 > > sentry-hdfs/sentry-hdfs-service/src/main/java/org/apache/sentry/hdfs/ > SentryHDFSServiceClientDefaultImpl.java 798bbef98b29af7c885c5ea747d77d > 2dad6c1693 > > sentry-hdfs/sentry-hdfs-service/src/main/java/org/apache/sentry/hdfs/ > SentryHDFSServiceProcessor.java 395618aa7271b1b739a08f98d284f12c37d31a7c > > > > > > Diff: https://reviews.apache.org/r/59895/diff/2/ > > > > > > Testing > > ------- > > > > > > Thanks, > > > > Sergio Pena > > > > > >
