On Mon, Aug 1, 2011 at 11:18 AM, Magnus Manske magnusmanske@googlemail.comwrote:

In case you missed it all the way over there in Haifa:-)

http://dirkriehle.com/2011/07/29/technical-report-on-wom-an-object-model-for...

Their model, while interesting and an excellent reference, makes some explicit choices that diverge from what we're currently working on:

* ugly but common structures where eg tables are opened/closed across templates are not supported * not all input is representable

They're not awful decisions -- and we might have made them 8-10 years ago had anybody made an attempt to *plan* the markup language. ;) But we have an existing data set of millions of documents that we have to support, and for the first next-generation parser I'm hoping to basically define something that's *very close* to how the current parser works, so that that first decade of Wikipedia documents can be fully used with a specified parser anytime in the future.

We can make the structures cleaner later and deprecate the old tables & whatnot -- parser functions and such allow for beautifully nested structures and a future wysiwyg world will take most of the low-level *markup* out of normal editors' faces -- but for now we have to make it work with what we've got. ;)

-- brion