Required: 4y degree, modern C++, a love of open source

This would give developers a way to query info about tickets. 

The configuration reload feature doesn't cover any use cases with multiple instances.
Given that, it is desirable to add this to irods-grid so that it may be invoked across the zone, something like `irods-grid reload all`

Davrods does not hit the actual put API, but performs an open/write/close. Please see this comment for details on which dynamic PEPs are engaged in the put operation: UtrechtUniversity/davrods#18 (comment)
The dynamic PEPs respond to actual API calls whereas the static PEPs are placed wherever the developer feels like it. In this case, the "oprType" can be set and rsDataObjClose understands this as having been a "put" and so the static PEP fires in a hard-coded location.

Providing such a feature would give administrators a way to see where time is spent regarding their policy and make informed decisions about how to improve it.

The iRODS repo does not provide any guidance on how to compile iRODS from source.
The project needs to introduce a new markdown file that explains how to compile iRODS. The markdown file should cover the following:
    Dependencies
    CMake options (e.g. CMAKE_INSTALL_PREFIX, CMAKE_BUILD_WITH_INSTALL_RPATH)
    iRODS options (e.g. IRODS_UNIT_TESTS_BUILD)
    Normal build steps (i.e. cmake . && ninja package)
    Recommended way of compiling iRODS (i.e. irods_development_environment)
The top-level README file can point to this new markdown file.

The client string printed by ips does not currently include any version information.

if you do a move and rename at the same time, for example,
'imv file1 coll1/file2', it will normally succeed if there's no conflicting
data-object name in the source collection (file2) but fail (giving error
CAT_NAME_EXISTS_AS_DATAOBJ) if there is, since, internally IRODS is doing
a rename and then a move. Please handle this by running multiple separate
'imv' commands.

Doing this will create a problem for iput --metadata. See #6409 for details.
This change in behavior is not good for existing deployments that rely on duplicate AVUs producing an error.

It's not uncommon to give another user access to data they don't own. There are two ways to do this:
Create a ticket for the target object and give the ticket to the user
Or, modify the permissions on the target object and then on the parent collection
The point of interest here is that using a ticket doesn't require modifying permissions on the parent collection. Great. The user can't look around the parent collection. All they can do is access the data associated with the ticket.
Q. Should modifying the permissions on the parent collection be necessary?
Q. Should modifying permissions on the target object result in the same behavior as using a ticket?

The rename API currently allows overwriting when FORCE_FLAG_KW is passed via the condInput of the dataObjInp_t.
imv does not expose a way to use this feature. We should add a -f option to allow for overwrites as the API already allows for.

icp has the ability to specify a destination resource via the -R option. It does not have an option to specify a source resource or replica number for the source data object. For many iCommands, this is -S for source resource and -n for replica number.
As with other iCommands, -S and -n should be considered incompatible parameters.
The recursive flag -r should also be considered while implementing this feature: What happens if the specified source resource or replica number is not valid for every data object in a collection?

Observed behavior (including steps to reproduce, if applicable)
The resource specified by -R is ignored and the existing replica on ufs0 is overwritten:

Investigation / Enhancement
The goal of this effort is to provide a binary encoding that is maintainable, efficient, extensible, versioned, documentable, and accessible to multiple programming languages.
We can achieve this by introducing pluggable binary encoders and decoders.
Potential Serialization Libraries
    Flatbuffers
    Cap'n Proto
    Protocol Buffers
    Cereal
    MessagePack
    Simple Binary Encoding (SBE)
Things to keep in mind
    Backwards compatibility
    Federation
    Encoding negotiation

This is about categorizing iRODS error codes so that icommands and other clients return meaningful POSIX error codes instead of a random integer between 1 and 256.
This can be achieved using a map and a function. Bonus points for allowing other programming languages (e.g. Java, Python, Go, etc.) to consume the existing C/C++ mappings in a convenient manner.
Reference material:
https://man7.org/linux/man-pages/man3/errno.3.html

The problem seems to be this: As of 579d0c2, delayed rule deletion is now correctly checking the permission level of the client user rather than the proxy user. The proxy user is always (or almost always) the service account rodsadmin account, which means that user will have permission to remove the delayed rule. Now that we are checking the client user, the usernames don't match because remote users always have the zone name appended but the delayed rule information is dropping the zone name from the username.
The solution, I think, is to write down the user name with the zone name in the delayed rule information in the catalog, and adjust the username check to account for this.

iRODS does not support changing the home collection during setup at this time. The setup script assumes certain things will be configured a particular way. This allows us to make assertions regarding correctness. After setup completes, you're free to change the home collection. To do that, you'd do the following:
    Create a new collection using imkdir.
    Open /var/lib/irods/.irods/irods_environment.json and set irods_home to the absolute (logical) path of the new collection.
With that said, allowing this in the future is a possibility.
Q. Is this something you need?
Q. Can you explain the motivation behind this?

this is a known issue with itouch on the S3 resource plugin. We will look into this.

iCommands should handle SIGINT and disconnect from the server before aborting, if possible. This will prevent connections from staying open unnecessarily after the user has aborted the iCommand.
To be clear, connections are cleaned up by the server after the TCP connection times out, but this just handles the client-initiated terminations more gracefully.

If you're using icommands, replace the use of "PAM" in your irods_environment.json file with "pam_password".

Further investigation has revealed that this only happens with 4.3.0.0 of the audit rule engine plugin installed and configured according to the training slide deck.
As this does not happen with a default configuration or with packages built from tip of main, I think this is actually an issue with the audit rule plugin and was fixed with irods/irods_rule_engine_plugin_audit_amqp#99.

PRC has historically done a query to see if the collection component of the requested new data object exists, and of course the Genquery is unable to distinguish between a collection it cannot write and one that doesn't exist.

During a large PRC put() both catalog and resource servers have normal (< 5% increase in both CPU and RAM) resource usage, but during a large PRC get() the catalog server's resource usage spikes strongly (nearly 100% CPU and RAM/swap) while the resource server's usage remains low. The spike is quicker with more threads and the transfer rate for get() is < 20% that of put() for the same data.

The configuration value for enabling PAM in 4.3.0 has changed from "PAM" to "pam_password".

EL7 (Centos 7, RHEL 7, and friends) needs a python3-pyodbc package. All other distros we support provide this already.

It is cumbersome not to be able to update my data files after they have been published.

To prevent errors, it would be good to have a front end check against the standard orcid format (e.g. 0000-0001-2345-6789).

I have a long list of categories in Yoda@WUR. When adding groups to a category all the way down, I have to scroll all the way up to add people and / or see group properties.

When typing in a tag and you hit enter once sometimes twice it will remove the tag entry

Current Behavior
Access Level (https://guidelines.openaire.eu/en/latest/literature/field_accesslevel.html) is sent to DataCite in the rights field
Expected Behavior
License URI is sent to DataCite in the rights field

When a data package is submitted to the vault. The metadata of the collection is copied over, but the metadata of individual data objects within the data package are not copied along side with the data objects.