Having a tool building a ZIM using the Mediawiki API is essential for people not having access to the system running Mediawiki.

For people with access to the system. using the API to get the content is not the best solution because it impeaches or makes difficult any post-treatments.
In addition, this is not the best approach to transmit all over HTTP if you have access to the content directly.

So, I think:
* A zimwritermw (zimwriter using the MW API) console tool should be developed (Tommi already started).
* Pediapress ZIM solution should be based on it ; otherwise I do not see how we will avoid duplicated work.
* zimwriterdisk (zimwriter using static HTML&Media files) should be continued (Tommi already has a stub I think)
* MW DumpHTML extension should be revisited to work correctly, with zimwriterdisk and potential additional post-treatment tools, this could be really powerful.

All this 4 tools separately are pretty simple to code and with all of them we might do everything we need.

So, I feel me concerned that neither DumpHTML nor zimwriterdisk are part of the discussions.