Thanks for your quick reply. I really don't know so much about XML but shouldn't it be just a matter of mapping the EverNote categories to the Zotero categories? Maybe it would even be posssible to use something like the XML Translator Generator? When I have some more time I will take a look at the developer wiki, and see if I can contribute in any way.

The structure of the EverNote XML files is fairly simple as far as I can tell. The whole thing is enclosed in <EVERNOTE> tags, within that there's three main categories, <NOTES>, <CATEGORIES> and <TEMPLATES>.

<EVERNOTE> has just one attribute: date, with the last change date given in YYYY/MM/DD hh:mm:ss format.

Every note is enclosed in <NOTE> tags which has the following attributes:
- created, again in YYYY/MM/DD hh:mm:ss format
- flags, values are "HasUnknownClip", "HasDropImage", "HasWebClip"
- id, for example [72EDBBD4-19B3-4E51-80C1F20AE70C817D]
- lock_content, value is either 1 or flag is absent
- short_height, with any numerical value [since this is a display variable it can be safely ignored in the translation]
- updated, in YYYY/MM/DD hh:mm:ss format
- history, in YYYY/MM/DD hh:mm:ss format
- history_root, for example [B75B22AF-2933-414B-9822E0CDD42F1A2E]]
- status, the only value I could find for this is "history"

Within <NOTE>, the actual content is enclosed in <CONTENT> tags, which has one attribute: type. The value is usually "html".

Within <NOTE>, the history of the note is enclosed in <HISTORY> tags. <HISTORY> contains a <NOTE> with one attribute: idref, for example [01FD65D7-F7AC-4BBD-A28554201F770B8E].

Within <NOTE>, the categories (tags) of the note are listed, enclosed in <CATEGORIES> tags.

Each <CATEGORY> has two attributes:
- idref, for example [F67298BB-DD62-4C2E-838962CD79B38FFD]
- name, which can be any text value
An optional third attribute is type, the value is usually "filter" [this is a filter for auto-tagging, which can be ignored in the translation]

Within <NOTE>, the resources of the note are enclosed in <RESOURCE> tags, usually images. They can have the following attributes:
- format, which is usually "base64"
- href, which can be either a local path or a URL.
- index, with a numerical value

This is followed by a great amount of gibberish, along the lines of "iVBORw0KGgoAAAANSUhEUgAAAoAAAAHgCAIAAAC6s0uzAAAgAElE ...", I assume this is the image in hexadecimal (?) format.

The notes section is followed by the categories section, which is enclosed in <CATEGORIES> tags.

The first category is the root category, it has an id such as [A6467135-1A5C-496C-A941D46D66F46E7B], and its name is "ROOT_CATEGORY"

within the root category, all the categories are listed as ref categories, enclosed in <REFCATEGORIES> tags. Each <REFCATEGORY> has an idref such as [E9CB2C2F-3E39-43DE-99B8ACDC47403367] and a name.

Each <CATEGORY> can have the following attributes:
- created, in YYYY/MM/DD hh:mm:ss format
- id, for example [A6467135-1A5C-496C-A941D46D66F46E7B], this is the same as the idref in the corresponding <REFCATEGORY>
- name, which can be any text value
- flags, values are "filter_saved", "filter_saved expanded"

Categories can have sub-categories, these are enclosed in <SUBCATEGORIES> tags, each <SUBCATEGORY> has an idref and a name.

Categories can have <REFCATEGORIES>, these are enclosed in <REFCATEGORIES> tags, each <REFCATEGORY> has an idref and a name.

Categories can have <PROPERTIES>, these have the following attributes:
- assign_category, value is 1, or flag is absent
- show_on_notebar_if_autoassigned, value is 1, or flag is absent

Properties can have filters, which are enclosed in <FILTER> tags, they have a type attribute such as "with_keywords", but since they all pertain to the EverNote filtering process they can be ignored in the translation. The content of the filter can be a list of keywords, a date, etc.

The categories section is followed by the templates section, which is enclosed in <TEMPLATE> tags. These only determine the assignment of little icons to specific notes, so I think this can be safely ignored in the translation.

For what it's worth, here's a brief description: <TEMPLATE> has created, id and name attributes.
<TEMPLATE> contains <CONTENT>, its type is usually "text".

<TEMPLATE> contains <RESOURCES>, the <RESOURCE> format is usually "base64", its other attributes are href and index.

That's more or less it, hope this helps. You can find a sample file of an EverNote export at http://playability.de/ENExport.xml

The Evernote XML Export format DTD can be found here:
http://xml.evernote.com/pub/evernote-export.dtd

While EverNote date isn't typically bibliographic in nature, a quick skim of the DTD and my afternoon of working with evernote doesn't reveal any data formats not currently supported by Zotero.

It looks to me like even the complex media types are embedded directly into the XML, so they could be imported directly into the local Zotero library.

evernote != endnote