Dear Vladimir,

> When it comes to code-related changes, the reviews are vague, and it is
> really hard (impossible?) to find consensus.
I somehow got an idea that ripping out classes that could lead to a 
NoClassDefFoundError for existing users did not fit the definition of "binary 
compability" for the log4j2 committers. As much as I would love to rip the 
classes in question out, I must admit that doing so is not binary compatible.

And if I recall correctly, the request on 
https://github.com/apache/log4j/pull/17 was to separate build changes from the 
code fixes and start with a PR to fix one CVE only (and have that fix to be 
something else than removing a class) so that can be reviewed in reasonable 
time. And if I read between the lines well, the committers wanted to see viable 
PRs before doing infra work that you are (correctly) suggesting. But apologies 
for butting in if I got something wrong.

Best regards,
Andrew

Reply via email to