Comment 29 for bug 1335323

Robert, you're right that if missing-component handlers are supposed to provide the component within the handler to be used in the same cache session, we shouldn't do negative caching, and/or we should have the restart reinitialize-source-registry-and-retry blow up the cache in addition to reinitializing the source-registry (and we should rename the restart, though we might have to keep the old name for compatibility's sake).

Once again, I'm not familiar with code that uses this hook, and would prefer if Zach were to tell exactly what hooks he needs into the system, rather than kluge his way around bugs and missing features. We fixed ASDF for the sake of Quicklisp in the past (see 2.015.2), we can do it again. We certainly did add hooks to ASDF for the sake of other extensions (e.g. POIU, ASDF-ENCODINGS, ASDF-FINALIZERS, ASDF-BUNDLE, etc.). There's nothing wrong for Quicklisp to depend on a recently fixed ASDF and provide it, just like it did with 2.014.6 then 2.26.