> On Feb 16, 2022, at 9:28 AM, Zowalla, Richard > <richard.zowa...@hs-heilbronn.de> wrote: > > The list archive does not contain any useful information regarding sxc.
Note the list archive frustratingly searches only the last month's worth of mail by default. Try this link: - https://lists.apache.org/list?dev@tomee.apache.org:lte=20y Here's a couple threads with information: - "SXC" February 6, 2019 https://lists.apache.org/thread/6h40wo6vfgrspjt2vzzfkpm8k2w42ym4 - "XML performance improvement" December 10, 2012 https://lists.apache.org/thread/f5bnv2nz917pjz8zyh7ktx6ngtysgron Here's the source code: - https://github.com/dblevins/sxc From the 2012 email, the key part from the last link, "We load 5,000 - 6,000 classes when TomEE boots and nearly 4,000 are JAXB generated." My memory is that this was over 50% of our start time. As a reference, generating classes in advance is basically one of the key things that makes Quarkus fast and the SXC/generation optimization a significant part of why we were faster to boot than everyone else in the pre-Quarkus days. The concept is valid and we definitely need to do significantly more of it, but there is a valid question on if this codebase is worth investment. I don't think it's worth significant investment, but I don't think we should remove it until we can replace it with something equivalent or better. If I had to do this again, I'd likely investigate if there's some way to do this with extending JAXB via the TomEE Patch Plugin. My advice would be to bytecode translate it for now if we can. -David
smime.p7s
Description: S/MIME cryptographic signature
