Note that I totally support the goal of not making any attempt to download these DTDs from the provided SYSTEM identifiers, and the idea of making sure that the character entities are always defined for documents specifying any of these PUBLIC identifiers.

I'm just:

1. Nitpicking about the wording here, since the DTD you "link to" here is quite different from the ones normally named by these PUBLIC identifiers.

and:

2. Suggesting that it could be reasonable for a user-agent to instead use a complete copy of the appropriate DTD that it happens to have on hand, perhaps looking it up in a catalog file.

I was going to suggest it might make sense to treat SYSTEM identifiers corresponding to these PUBLIC identifiers in the same way, but then I noticed that my system's XML catalogs don't do this (except for identifiers starting with "http://www.w3.org/TR/MathML2/dtd/"), so it probably isn't useful.

(I may have temporarily forgotten that SYSTEM identifiers are mandatory in XML. I really would prefer that they be optional: the security, document stability[1], and performance implications of actually dereferencing these being what they are, one would often prefer to rely entirely on catalog files; having made this choice, mandatory SYSTEM identifiers are nothing but a nuisance -- one must either use bogus identifiers, or be careful to always disable networking, lest something not in a catalog sneak in ...)

[1]: Not everyone can keep a URL alive forever like the W3 or OASIS; even docbook.org makes me a bit nervous.

The term "additional information" is from XML:

# An XML processor attempting to retrieve the entity's content may use any 
# combination of the public and system identifiers as well as additional 
# information outside the scope of this specification to try to generate an 
# alternative URI reference.