Any chance this is a solved problem? What happens in the JDT/Corona toolkit when they need to go validate schema against things like the WSDL spec? Or am I thinking a little sideways, here?

Looking through my DTD-loading glasses, your stuff has to be chrome.

If you're only working on one particular external package, I'd register that with a custom protocol (instead of jar:) and make that protocol cache and backup to whatever. It's likely to be a tricky mess, though.

(From Alex: Oh, I wish it were quite that simple. I'm trying now to implement a protocol.)

#3 sounds like you're implementing a RAMdisk in XPCOM. The thing with that is you'd only need it on systems that don't have their own RAM filesystem and you could use the native one elsewhere. Sounds like a better idea to me.

(From Alex: When you get down to it, the third option is a lot like what you suggest. The big difference is that this RAMdisk is much less complete, and goes away at the end of the Mozilla run. Besides, we have no way of knowing if the user has a RAMdisk available or not. So I'm going on the assumption that Verbosio should try to have its own system.

I said try, not succeed. But I am making progress... the data: protocol handler and channel implementations are really close to what I want to do.)