xml queueing design help

9.2 on windows (development)...moving to Linux in a month. Production will be linux.

  I've just begun work on a document submission application. There will be 3 seperate types of documents submitted. We are going to make use of AQ. The documents need to be processed in a certain order so we are going to use priority queueing. One of the parameters into our submit_queue procedure will tell us what type of file is being submitted. 2 of the file types are highly variable in structure but small in size. The third is simple measurement data that is pretty constant but is very large. The people originally working on the system envisioned that all 3 documents will be xml because they will be passed around among various platforms. I don't have a problem with that on the 2 smaller, more variable documents but i'm worried about processing performance of the larger file.

Currently, the plan is to submit the file into the queue as a clob and then we'll use xquery to move the data into relational tables where it will be evaluated by a qa engine developed in .net (no control over that, they want to reuse code already in place). We'll be using clobs in the queue because there are fairly strict time constraints for notifications that the submission was received and storing it as an xmltype will slow that process down.

once it is in the queue, i'm worried that moving the xml data to the qa area is going to be s-l-o-w. my initial thought was to test that against placing the submission files on the server and then just using the queue as a notification that the file is there. We would then use external tables to move the data. The only problem is that i've been told we flat out cannot do that. The files must go directly into the database. The next option is strip out the xml tags on the app server and submit the file as a fixed field text document with an automatically generated xml document describing the format of that text document. This will Greatly reduce the size of the file but i'm not sure how fast i can get at clob data.

So, i guess my questions are:

Any other options i'm not thinking of?
Anyone go through a similar project? If so how was the performance? How slow is xquery? (i can answer this myself but i thought i'd throw it out there and possibly save some time) How slow is sql access to fixed lenght clob data?(i can answer this myself but i thought i'd throw it out there and possibly save some time) ..oracle text would help me here correct?

Thanks for any help/comments!
chris

--
http://www.freelists.org/webpage/oracle-l