Is Docbook Assembly easier to process than standard Docbook?

[Lech Rzedzicki <lech@kode1100.com>]

I'm sure there's good reasons for DB Assembly to be the way it is right now, but I need to understand those reasons.

If I need to read or modify, let's say the first para of a topic, I have to parse the <structure>, fetch the id from the the module, then parse the resources to fetch the URI of the actual file and finally fetch and parse the topic and grab the first para.
Had this been one big file, it'd simply be current-topic/para[1].

Is anyone else doing Component Content Management System which chunks the content, but leaves it up to the application to chunk it?
I find the whole split into topic/map a bit artificial, as typically even a topic is a sort of map of its content and very often superstructures like a chapter need to be treated as a single object.

What are your experiences and what are the best practices for the two scenarios described above:
1. Treating a sub-map or a map as a single object
2. Fetching/reusing a portion of the topic?

Regards,

Lech