Re: [agentic-ai-dev] Maven Coordinates/Package Names

[Reza Rahman <reza_rahman@xxxxxxxx>]

I don’t think there’s really a permissions issue of some sort. I suspect Ivar would have told us if there was.

It’s just a semantic issue. I think just one simple group ID for all things AI will be easier for end developers. It’s not like it’s that segmented of a community in the end. It’s what a lot of OSS projects tend to do with closely related artifacts.

I suggest we just go with a simple group ID and see what the spec committee says. jakarta.ai.agent as a group ID wouldn’t be the end of the world if they say we have to do it that way, just a tad more verbose than really necessary.

> On Mar 8, 2026, at 4:30 AM, Piotr Żygieło via agentic-ai-dev <agentic-ai-dev@xxxxxxxxxxx> wrote:
> 
> Hi List
> 
> Seeing so many thumbs up gives me hope someone will be able to help me
> correct my understanding of the case.
> It's only about Maven coordinates (groupId to be specific) for me here.
> 
>> On Sun, 8 Mar 2026 at 00:38, Reza Rahman via agentic-ai-dev
>> <agentic-ai-dev@xxxxxxxxxxx> wrote:
>> 
>> * Group ID: jakarta.ai (this will enable us to place all future AI work
>> into one neat place such as jakarta-llm or jakarta-mcp along with
>> jakarta-agentic-ai)
>> 
>> * API coordinate: jakarta.agentic-ai-api
>> * Spec coordinate: jakarta-agentic-ai-spec
>> * TCK coordinate: jakarta-agentic-ai-tck
>> * Examples coordinate: jakarta-agentic-ai-examples (examples included in
>> the project itself like most OSS projects)
> 
> What I think, which every single point I might get or assume wrong:
> 
> 1. Specifications within JakartaEE have separate Sonatype namespaces,
> and access is restricted in the way that spec X is able to publish to
> its namespace but not to the namespace assigned to spec Y. (General
> example: Annotations - jakarta.annotation and Authorization -
> jakarta.authorization).
> 
> 2. Today it's not possible to deploy to more than one namespace in
> Central Portal in one go.
> 
> 3. New AI related spec - let's say Z - would be new spec, like Jakarta
> Z AI. Could use Agentic AI, could not do so. Separate /who list,
> separate goal, etc.
> 
> 4. I assume "Z AI" is willing to be in jakarta.ai groupId and to
> follow API, Spec, TCK naming.
> 
> To get that - all proposed above agentic-ai would be on the same level
> as "such as jakarta-llm or jakarta-mcp".
> Hence for me it looks that all agentic-ai artifactIds shall be
> separate namespaces, right?
> 
> What if the project would need additional artifact (like
> websocket-client-api does)? Separate namespace :(
> And all "Z AI" namespaces... And others...
> 
> That's why I think there could be an additional level in groupId to
> partition the namespace now, to keep deployment simple, to allow
> anything to happen within that as the project needs changes, without
> the assistance from EF helpdesk and Sonatype wrt to multiple
> namespaces.
> 
> Just like there is for example jakarta.xml.ws, jakarta.xml.soap and
> jakarta.xml.bind.
> 
> Please consider updating groupId to something like jakarta.ai.agentic
> to create this structure and not put all anticipated future specs into
> the same bag.
> 
> --
> Piotrek
> _______________________________________________
> agentic-ai-dev mailing list
> agentic-ai-dev@xxxxxxxxxxx
> To unsubscribe from this list, visit https://accounts.eclipse.org