There's syslog the protocol, syslog the API, and syslog the daemon. There are syslog daemon and protocol replacements that are much better, but retain the API for wide-spread compatibility.
I think my contention would be that if you shake off the early-90s-Unixisms and adopt any modern store (forget Redis even and just "use sqlite instead of fopen"), the machinery required to do 98% of what the best syslog daemons do is trivial. The syslog infrastructure just doesn't add much... except compatibility with a very dubious ecosystem of tools.
