sunyuhan1998 commented on PR #10895:
URL: https://github.com/apache/gravitino/pull/10895#issuecomment-4412139877

   Thanks @yuqi1129 for the review! Replying to your questions:
   
   ---
   
   **`CredentialPropertyUtils.java:55`** — Those values are only used for 
Lance. Is it more proper to move them to the lance module?
   
   These constants and the `lanceCredentialPropertyMap` / `toLanceProperties()` 
are indeed Lance-specific, but `CredentialPropertyUtils` follows a centralized 
pattern where all table format credential transformations live in one class. 
The Iceberg equivalents (`ICEBERG_S3_ACCESS_KEY_ID`, `toIcebergProperties()`, 
etc.) are in the same class. Keeping them together seems consistent with the 
existing pattern. Would it be acceptable to keep it as is, or do you think 
moving them to the lance module would be better?
   
   ---
   
   **`LanceTableOperations.java:100`** — So, by default, we will get the 
credentials in the storage options?
   
   No, credentials are only injected when the client explicitly sets 
`vendCredentials = true` in the request. When `vendCredentials` is not set or 
is `false`, `credentialPrivilege` remains `null`, and the code falls into the 
`else` branch which returns only the original storage options without any 
credentials. See the `if (credentialPrivilege != null)` check in 
`GravitinoLanceTableOperations.describeTable()`.
   
   ---
   
   **`LanceTableOperations.java:117`** — Why is it `CredentialPrivilege.READ` 
when the length of `parts` is not 3?
   
   The `parts` should always have length 3 (catalog/schema/table) in normal 
flow. When it doesn't, using `READ` is a conservative safe default — it's 
better to under-provision than over-provision credentials. That said, this 
branch should never be reached in practice since upstream validation already 
enforces the 3-level structure. Would it be better to throw an explicit 
exception here instead?
   
   ---
   
   **`LanceTableOperations.java:126`** — Is this to check whether the current 
user has the write privileges? How can we get the current user?
   
   Yes, `MetadataAuthzHelper.checkPermission()` checks whether the current 
authenticated user has write privilege on the table. The current user is 
obtained from the Gravitino authentication context — when the request arrives, 
the identity is already established and bound to the thread context. During 
credential generation, `CatalogCredentialManager` calls 
`PrincipalUtils.getCurrentUserName()` to retrieve the authenticated username 
and generates credentials scoped to that user's permissions.


-- 
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.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to