> I get the impression that folks who haven't worked in such large > mono-repos aren't aware of just how big a multi-year effort it is to > move a repo like that onto a new JDK version. And that's just the VM, > even before you allow devs to change the language level and start > using the new features and libraries. That's just the two really big > mono-repos I have personally worked in. I'm willing to bet that some > other big Java shops are in similar positions.
Sorry that I made the wrong impression. I know this is a big effort from personal work experience. But it can be done. It must be done. I asked the CI team to run all CI builds with a parallel JDK 11 build, and they did that. This way, it was easy to see which project was not buildable using Java 11, so teams could work on that. For the apps I work on, my team uses all LTS versions plus the latest GA version (if not an LTS release). By the way, this was an excellent recommendation by a former IBM employee. ... and that was all done without raising the language level. Besides that, most big (tech) companies do not allow unmaintained or unsupported software. I am puzzled how this could work out with the state of the libraries you mentioned. There must be other violations in the first place, and someone allowed them to happen. If they hadn't been allowed to happen, there would have been no need to catch management attention. > Switching back and forth between JDKs for open source development is > doable but an unnecessary hassle. I am puzzled why this would be needed. All Java 8 apps I know were easy to build on Java 11 (with release=8) after only very few adjustments. That said, I do not think Apache Maven should wait for two companies with mono-repos to update their technical debt. It is just not justifiable for all the other users. I agree with Gary that an EOL schedule might be our best shot here. Am Mo., 5. Feb. 2024 um 19:56 Uhr schrieb Elliotte Rusty Harold <elh...@ibiblio.org>: > > On Mon, Feb 5, 2024 at 12:01 PM Benjamin Marwell <bmarw...@gmail.com> wrote: > > > Why 17? 11 is often earlier EOL'd than 8 and 17, so I see absolutely > > no advantage of going to 11: > > > > The advantage of going with 11 instead of 17 is that at least 2 really > big tech companies I could name (and who you can probably guess from > my linked in) have only recently completed their own migration to Java > 11. At least one of those two companies might still employ a PMC > member (though I haven't checked post-layoffs), maybe more than one. > Both have actively supported Maven development over the years, though > that support ebbs and flows depending on corporate priorities. > > I get the impression that folks who haven't worked in such large > mono-repos aren't aware of just how big a multi-year effort it is to > move a repo like that onto a new JDK version. And that's just the VM, > even before you allow devs to change the language level and start > using the new features and libraries. That's just the two really big > mono-repos I have personally worked in. I'm willing to bet that some > other big Java shops are in similar positions. > > Switching back and forth between JDKs for open source development is > doable but an unnecessary hassle. I've done it before, but even > switching from JDK 8 to 11 is an extra paper cut. It kills time every > time spotless fails simply because I'm using Java 8 instead of 11. > > Most importantly, it will be even harder to sell management on the > benefit of spending developer time on Maven 4 development, if it isn't > suitable for that company's own open source projects which, last I > checked, were still on Java 8. (OK, I just spot checked and the first > one I looked at is in fact still compiling for Java *1.7*, probably > because that's where their customers are). > > I'm thinking back to the projects I had to justify to management a few > years and one company back, and it definitely would have been harder > then if I had to tell them what we were contributing would only work > on Java 11 or later. Maybe today I could sell them on Java 11 (or > maybe not, if they aren't paying attention to Maven at all any more) > but Java 17 would be a non-starter. They might prefer to spend their > resources on a build tool they own, or maybe just not spend the dev > hours at all. > > tldr: every uptick in version requirements bleeds that many more > contributors out of the pool. > > -- > Elliotte Rusty Harold > elh...@ibiblio.org > > --------------------------------------------------------------------- > To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org > For additional commands, e-mail: dev-h...@maven.apache.org > --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org For additional commands, e-mail: dev-h...@maven.apache.org
