> To be clear, if it will be Option 1 (new package), that will happen on
> build time as well (most likely).

If you intend to go for option 1, that is a sensible decision.

Am Do., 9. Nov. 2023 um 12:16 Uhr schrieb Daniel Dekany
<daniel.dek...@gmail.com>:
>
> > Furthermore, its automated with Maven Shade plugin, which prevents code
> duplication and maintenance headaches.
>
> To be clear, if it will be Option 1 (new package), that will happen on
> build time as well (most likely).
>
> On Thu, Nov 9, 2023 at 7:14 AM <le...@flowlogix.com> wrote:
>
> > As a Jakarta EE, PrimeFaces and Apache Shiro committer and contributor,
> > there is only one option, which is #2. This is the “standard” way
> > countless other projects added support for Jakarta EE 9+
> > namespace conversion, and FreeMarker should be no different.
> > Option 2 is clean, standard and everybody understands how to do it.
> > Furthermore, its automated with Maven Shade plugin, which prevents code
> > duplication and maintenance headaches.
> >
> > On 2023/11/07 22:50:02 Daniel Dekany wrote:
> > > The package of Servlet related classes has changed because of Jakarta,
> > > which breaks our Servlet support (freemarker.ext.servlet), which is
> > packged
> > > into freemarker.jar.
> > >
> > > We have to choose which end result we want (ignore the "how" for now) as
> > > the solution, from these two (as far as I can tell):
> > >
> > > 1. We can copy the `freemarker.ext.servlet` package into
> > > `freemarker.ext.jakartaservlet` (or such), and we will only have the
> > normal
> > > artifact in Maven Central, which contains that, and also the older
> > > freemarker.ext.servlet. Explanation: As you probably know, 2.x has a
> > single
> > > monolithic freemarker.jar artifact, which already contains support
> > classes
> > > of various optional dependencies. We already support multiple
> > incompatible
> > > Serlvet/JSP versions, and has separate version-specific classes for some.
> > > But, classes like  freemarker.ext.servlet.FreemarkerServlet managed to
> > stay
> > > common amongst Servlet API versions. For the Jakarta change not even that
> > > can remain common of course.
> > >
> > > 2. We can have an additional artifact variant (let's say via Maven
> > > classifier "jakarta"), that still uses the `freemarker.ext.servlet`
> > > package, but there that links to the Jakarta Servlet classes. This
> > artifact
> > > will drop support for pre-Jakarta Servlet/JSP versions.
> > >
> > > Possibility 1 pro: We don't have to publish one more artifact. Also, then
> > > users don't have to fiddle with dependency management to choose the
> > > artifact with the "jakarta" classifier.
> > >
> > > Possibility 1 con: Any existing dependent Java code that used
> > > `freemarker.ext.servlet` so far, and wants to migrate to a Jakarta
> > Servlet
> > > container, has to be modified to link to `freemarker.ext.jakartaservlet`
> > > instead. That sounds quite bad, however, the same dependent Java code
> > > likely has to be modified anyway, to link to Jakarta Servlet classes.
> > > Except, there are tools, like
> > > https://github.com/apache/tomcat-jakartaee-migration, that transforms
> > jar-s
> > > to depend on Jakarta Servlet API, but same tools of course won't replace
> > > links to freemarker.ext.servlet with freemarker.ext.jakartaservlet, so
> > some
> > > pain is expected. Also, `web.xml`-s that refer to
> > > `freemarker.ext.servlet.FreemarkerSerlvet` also have to be modified, if
> > > someone uses a Jakarta container.
> > >
> > > Opinions?
> > >
> > > Note 1: We had two attempts so far on this issue, but certainly the
> > actual
> > > solution will be a 3rd one. Anyway, the "how" is now not the point now,
> > but
> > > here they are:
> > >
> > >   - https://github.com/apache/freemarker/pull/94
> > >   - https://github.com/apache/freemarker/pull/95
> > >
> > > Note 2: At some later(!) point, maybe in a FreeMarker 2.4.0, we can get
> > rid
> > > of non-Jakarta servlet support. At the same point, we will also get rid
> > of
> > > the GAE/non-GAE variety. So we could end up with just a single variant of
> > > the freemarker 2.x artifact, over time.
> >
>
>
> --
> Best regards,
> Daniel Dekany

Reply via email to