I won't discuss the accuracy of the model you purposed because, obviously you're better informed than me about real life requirements, works, and works not in the bliki world but some questions came up to me.

~~
When coming up with a domain model you need to define a certain granularity threshold so you don't end up modeling words, characters, bytes, etc. Do you have a specific metric for this case, that is, have something like "That concept is too low level" or "That's over engineering" while drawing the model? (Note that I'm not suggesting you over engineered, on the contrary)
~~

That was just a curiosity. About the domain models, why isn't post a wiki page? Are you already considering presentation issues here? "Converting" the concepts of the entry/meta data model into web pages requires a whole sub model do be described.

Other thing that came up to my mind what that you are sticking with "... and then we have the Blog, which is not an entry, that is a collection of posts..." and probably you live better with a generalization of what blog is. Imagine a vwiki. A site like youtube were all textual content are wiki fragments and the video is probably an attachment. Or a swiki. A social site were peoples profiles are wiki pages and relation between people are made through wiki links, labels, comments, etc. Or even a wikizilla. A bug tracking system were issues are wiki pages with comments and attachments and all (actually Trac is such a beast).

~~
While talking about blog's concept generalization an epiphany about wiki fragments and composition was forming but never mind it's gone :-)
~~

What I'm saying is that generalizing what Blog is in the entry model may bring a lot more potential, better prepare you for the presentation issue and generate a bag of new questions/problems.

PS: I can never remember snipsnaps's wiki syntax, bah