question about hive.allow-register-partition-procedure

[DongSeungLee]

i have a question, why allow-register-partition-procedure config is false as default, otherwise
unregister single partition is available.

The official document says that security problem might occur.
What types of security problem can occur?

you can see related PR(allow-register-partition-procedure)
d14ed68

[mosabua]

I assume because the location already exists and presumably contains data, and that can potentially lead to issues like data corruption when write operations are performed. But I am mostly guessing - @findepi @kokosing might have more info

[findepi]

I assume because the location already exists and presumably contains data

this can lead to inadvertent data exposure (unprivileged data access)

[DongSeungLee]

let me guess, you're saying that if someone does not know the details of HDFS(s3) path, using register partition procedure, someone can deduce the structure of stored data.
am i correct?

[hashhar]

Someone can register the partition in their own table which points to data from some other table that they don't have access to. So even if user can't query the table directly they can create their own table and add other tables partitions to it and query the data that way.

[DongSeungLee]

i appreciate your answer. It is very helpful!

[lozbrown]

@hashhar @findepi wouldn't URL authorization similar to HiveServer2 be the answer here? IE granular control of locations users can/can't create tables/partitions