The logger in iRODS hard-codes the value of log_facility to local0. This property and value are not important to the operation of iRODS and should be removed.

While the client api allowlist's functionality will remain the same, we should change how it is held to match other modern iRODS singleton interfaces.
This is more about interface consistency than anything.

See #6624 for more details about how we got here...

Okay, so the tasks here are:
Add mention/description of LOCKED_DATA_OBJECT_ACCESS and INTERMEDIATE_REPLICA_ACCESS in Troubleshooting for "Data object stuck in locked (or intermediate) status"
Add new page for columns available to iquest (aka iquest attrs)

ibun supports compressed structured files, but I have found no microservices that similarly do so. All I see is msiTarFileCreate/Extract and they don't support compression. Creating a generic structured file microservice pair that supports compression and data type is one idea, another would be to have a separate microservice pair for each data type supported by ibun.

msiTarFileCreate contains a force switch, but msiTarFileExtract does not. It would be nice to have similar options for complimentary microservices. Also, some microservices use a dedicated parameter for this while others include it in a msKeyValStr. And the process status is sometimes returned both directly and by reference (msiTarFileExtract) and sometimes only directly (msiTarFileCreate). Consistency in these would also be nice.

msiExecCmd sets up communication pipes, forks, closes excess file descriptors and executes a program.
The file descriptors are closed by closeAllL1desc, which is called from nowhere else. It starts at L1 to discover the descriptors, and then goes back up to L3 to close them. However, this has all sorts of effects -- apart from closing the file descriptors -- that are unwanted at this point. It is not appropriate that pep_resource_close_pre is called because the relevant resource is still open in the parent process, and it's not appropriate to run any rules in a forked process.

Does anyone use RBUDP for data movement?
If no one relies on it, we should remove it.

We will support Ubuntu22.04 in an upcoming 4.3.x release. No known blockers at this time.
We will not support Ubuntu22.04 for 4.2.x. - see earlier discussion at #4883.

For long running agents, updating one or more resources means the resc_mgr needs to be updated as well. Not updating the resc_mgr leads to incorrect results. It also means iRODS may not be able to carry out certain operations because the catalog and resc_mgr are not synchronized.

The only documentation I see that speaks of those operators is here: https://docs.irods.org/4.3.0/icommands/metadata/#qu
We'll look into providing more documentation for this.
As for new operators, there are no plans to add new functionality to the existing GenQuery implementation. The new GenQuery parser we're working on will address these issues.

iRODS does not declare any dependencies on rsyslog or logrotate, yet setup includes logic that attempts to configure and launch them.
Given that there are multiple implementations of syslog, iRODS cannot assume all deployments will use rsyslog. Choosing which syslog implementation to use must be made by the administrator.
Therefore, all logic focused on setting up these services should be removed. Documentation should be provided that helps administrators with configuring syslog and logrotate.

I'm using python-irodsclient version 1.1.4 with iRODS 4.2.10. I use irods.manager.DataObjectManager.open to open a data object using open mode w and implicitly allow the data object to be created, If the data object already exists, the library appears to change the open mode to w+. This changes the access to the data object from write-only to read-write.

I've verified that the math is correct. We suspect the speeds reported by you are because the transfers are all local (no remote network involved) and the OS is being very smart.

iRODSSession.data_objects.put() doesn't refresh the state of the catalogue after/before each call.

Unfortunately put( ...) in PRC isn't quite the same as iput - one difference being it doesn't use the PUT api

Looking at this again, all we get as repr( )-style output is UNIX_FILE_WRITE_ERR(None,), and ....
We could have at least included the errno code (which was 28) and/or the symbol and strerror-message [ (ENOSPC) / 'No space left on device']. I'll make this an issue in its own right.

Setting a group permission on a collection exposes all users of the group. However this works normal for data objects.

This can be closed because I have learned that the keyword argument report_raw_acls is controlling that. Though I guess this comes already as True for data objects by default.

Querying acl on (getting permission of ) an object doesn't bring g prefix specifying group based permission although there is already an granted group permission exists. AccessManager behaves it like a normal user. On the other hand ils -A shows it is a group user.

Allthough metadata items can be queried and filtered for create_time/modify_time columns, the higher level class iRODSMeta does not expose these fields. Consequently, they are not available when fetching metadata from collections and data objects (and resources and users).

get( ) and other methods call logical path normalization implicitly, so this is a priority to fix.
the irods.path.iRODSPath carelessly uses os.normpath which on MSWindows applies the finishing touch of replacing any forward slashes ( / ) with the backward ones ( \ ) more standard on that OS.

packstruct tags which have been added to an API can be left in during client communications with older servers and they will be ignored (affects metadata , ticket APIs)

The affiliation field in the metadata form is a free text form that invites different types of spelling of affiliation (such as UU, Utrecht University; or WUR, Wageningen UR, Wageningen University and Research, Wageningen University & Research, Wageningen University, etc). Although a person identifier can be provided, an affiliation identifier is not present.

During the deployment with ansible we noticed that the scripts located in /var/lib/irods/msiExecCmd_bin were not the latest ones available in the Yoda github. For example "admin-tape-archive-set-state.sh".

Please make the user's 'domain' variable instead of uu.nl.