8:55 a.m.
On 9 Aug 2023, at 15:42, Eduardo Martins <emartins(a)redhat.com>
wrote:
> On 9 Aug 2023, at 12:19, Jeff Mesnil <jmesnil(a)redhat.com> wrote:
>
> On 9 Aug 2023, at 12:41, Eduardo Martins <emartins(a)redhat.com> wrote:
>>
>> Oh I didn’t meant we should only have one solution, was just trying to clarify
what you are targeting now.
>
>
> You’re right. That’s a good thing to clarify.
> My goal with that thread is to make WildFly more accessible and improve its user
experience.
> Part of that is the tooling (code generator, maven archetype, whatever) but if we do
not let users find out and help them get started, all the tooling improvements we are
doing are lost to them.
>
> In that sense, I don’t see why we could not create a getting started page *now* with
what we have.
My concern is more about us going forward with new archetypes, and in 6 months possibly
abandon such direction.
IMHO We should first learn why Quarkus team opted for a maven plugin instead of
archetypes, and plan to go forward on the right direction. If we should not go with
archetypes in the future then we could for now:
a) use the existent archetypes
As I mention, they are not a good fit to « get started ».
If an user needs to create a EAR, they are fine but this should not be the introduction to
WildFly…
We can discuss the webapp archetype though… Again, starting with JSF and JPA seems a big
jump in complexity. If an users does not need them (we can argue about JPA but there is a
bunch of backends apps that do not need JSF), we have to tell the users how to remove
things and that’s not how a good user story can progress.
b) use hello-world quickstart (going in a flow similar to quarkus get
started page. 1. curl command to download the Quickstarts, 2. Build hello-world with
wildfly server provisioning, 3. Run provisioned server, 4. Access the running app)
That’s not a great start imho. Once an user has done that, they are stuck with the
structure of the Quickstarts and will have to dp extensive work to get a « clean »
application from it (without deps to wildfly-quickstart-parent, cleaning up the openshift,
provisioned-server, bootable-jar profiles, etc.)
The Quickstarts are a good collection of examples but they are not great to get started…
> I don’t know. However the output of their code generator is a
Maven project. It would be the same for WildFly.
> We can start with a Maven archetype now (Darran’s draft PR is quite close to be ready
imho) and update it with a code generator when we have one. You mention you wanted to
start this in Q4 and I don’t know the amount of work to create such generator but I don’t
want to wait until Q1 2024 or later to improve the user experience of WildFly (and
wildly.org).
Totally agree we should improve WFLY users experience now (and all the time), just a bit
concerned that we may go in a direction that we may discard in a few months. Please note
that while developing and maintaining the archetype is not a big deal, we also need to
cover other aspects like web and docs.
That’s great to be able to discuss the direction from the user experience. Could you point
to the RFE for the code generator so that we can see what are the other choices that were
discussed?
Jeff
--
Jeff Mesnil
Engineer @ Red Hat JBoss EAP
http://jmesnil.net/