As stated in their 2003 Annual Report, the EPO stated: (http://www.european-patent-office.org/epo/an_rep/hmlfiles/epoque.htm)

We agree with this assessment 100%. This is why you can no longer have just the full text to read as you search to narrow down your list of candidate patents. You must also have the images of the patents as they are legally issued right in front of you with the text as you read patents from your search results. This is what the examiners use, and what is available at the USPTO, and this was the first major goal of P9. The searchers wanted to be able to have the text of the patent on the screen, and be able to flip through the pages of the legal documents (which is what patents are) while independantly scrolling the text. P9 provides this  - and it had to be fast - after all, to professional searchers, time is literally money.

Most of the searchers in the firm were examiners at the USPTO for a number of years, and therefore had used the EAST and WEST search systems at the USPTO extensively (both the public system, and private examiner's system, as well). The first phase of our design was to go to the USPTO and closely watch and document several professional patent searches from start to finish on the WEST system. From this we could see exactly what features they used the most, and why. We also looked closely at the WEST interface itself to see what interface features we wanted to implement (and could implement). Quite often, the learning curve is a big part of the cost of any new system, and the less they had to learn the better. For this reason, partially mimicking the WEST interface made sense, to an extent. After this visit, we sat down and built a list of "must haves" for the initial system, and finally we went into some detail on the user interface.  Here is a short list of the most important Must Haves that were decided on. The list is divided into 5 basic sections:

Indexing/search data:
1)  The system would have the full text of all US patents from 1976 to the current issue week, and the full text of all published applications from 2001 to the current publication week (indexed and searchable in the system). 
2) The system would have all patent and application high resolution image pages back to 1790 (literally all of them). The images with no matching text files would be indexed and searchable by document number and current U.S. classification and assigned current cross references classes.  Those of you that have ever "searched the stacks" understand what this is about.
3) The text would be divided into the following searchable/selectable indexes: Title, Abstract, Inventor, Assignee, Examiner, Legal Rep (Agent/Attorney/Firm), Government interest text, Parent case text, Foreign references, other publication references, Claims, and Detailed descriptive text sections.
4) P9 would be able to execute Boolean full text searches on any single index, or all of the text indexes simultaneously. You would also be able to build on previous search results in a multi-level fashion (e.g., Results of Search 1 logically ANDed with results of search2 (out to 1000 levels). Proximity operators would be avaiable to search simple phrases and words Adjacent or Near each other.
5) Additional searchable field-based indexes would include: Document number, U.S. current class, current Cross reference classes, Application number, Appl. date, PCT information, Application priority data (document listings), International class (as issued), and US patent references.
6) The system would have the ability to perform "one click" forward citation searches on U.S. patents (list all patents that reference this patent), and also reverse citation searches from 1 to 3 levels deep (e.g., this patent references 10 patents, and each of those patents reference more patents who also reference yet more patents... all returned as one set of search results). 

Hit-list requirements (display, sorting and paging):
7)  Search should be able to return and track up to 2 million active documents with 50 to 500 items per page listed at a time with basic "line item" details. 
8) The displayed portion of the Hit List  (50 to 500 at a time out of up to 2,000,000 returned in a search) would be sortable by Document number, U.S. Class/subclass, and Issue Date. 
9) The detailed portion must also be "slideable" up and down the total active hit list  (e.g., Next 100, Prev 100, Next 500, Prev. 500)  with the ability to re-sort this smaller group (the displayed portion) as desired. 
10) Ability to turn on and off additional details in the displayed section (50 to 500) of the hit list to include the  Inventor's last name, primary examiner, the international class (as published) and at least 100 characters of the title.

Online U.S. classification listing and other classification requirements
11) The hit list must be "properly sortable" by U.S. Class, which means numerically by class first, then, alpha-numerically by subclass. The USPTO WEST system (and 2 commercial systems we are aware of) do not sort class and subclass correctly (yes, this surprised us). They do a straight alphabetic sort (e.g., 23/8 is before 2/54). This annoyed the professional searchers quit a bit.
12) When the class/subclass listed by each displayed hit list item is clicked-on (mouse click), it must pop-up a properly indented, subject hierarchy classification listing (e.g., "dot indented") as assigned by the USPTO, in a scrollable, resizable windows. 
13) The current class information (Current U.S. Class and Cross Reference classes) on all documents and the pop-up class listings must be updated at least quarterly (4 times/yr) from the USPTO Class subscription data.

System Interface type and Text/Image display requirements 
14)  The system would work across the Internet or an intranet using a popular late model browser with large LCD Monitors (20 inch). 
15) Like the WEST system, and unlike the public USPTO Internet search system, image pages must be easily displayed in the browser with no special graphics plug-in. In other words, the system must convert patent image pages into native Internet graphics formats (on the fly) for high speed display (PNG, JPG or GIF).
16) Each of the patents/applications in the detailed section of the hit-list (the current 50 to 500 out of up to 2 million found in the search) must have a button (a link) that will display the full text (if available) in a dedicated window, and another link that will display the image pages (if available) in another dedicated window. In fact, any patent text displayed must have active links for the US references to allow instantaneous display of that patent's full text. 
17) A text only mode should be selectable where the searcher can click one button and display the full text of the next (or previous) patent in the hit list (e.g., continuously press next until all patents in the list have been displayed).
18) An image only mode should be selectable where the searcher can click one button and display the first image page of the next (or previous) patent in the hit list, and button that also move you to the next or previous page in the current patent image file.

Auxiliary functions for saving and managing list of documents and searches
20)  Searchers  must be able to create managed lists of documents with the ability to easily add or delete documents found in the hit-list to these managed lists (we call these lists collections).
23) Be able to easily order a "collection" of the documents in a format such as multipage TIF or PDF  delivered to the workstation where the browser is running (via a standard download link  FTP or HTTP for things like recording onto CD or DVD media, or to print for reporting purposes.).

P9 does all of this, and more. If you are an APLS user and you are interested in testing the system with us, giving us your opinions and answering some simple questions from time to time (via email), please call us and let us know. We will set it up and provide you with access and instructions for P9. Thank you! The number to call is 800-832-2823. Please tell anyone that answers that you are interested in testing the P9 Search System.