diff --git a/doc/persistent_ids.md b/doc/persistent_ids.md
new file mode 100644
index 0000000..b21bfe8
--- /dev/null
+++ b/doc/persistent_ids.md
@@ -0,0 +1,53 @@
+# Persistent IDs
+
+## Centralized DISC ID management for Dexpi Equipment IDs
+We suggest separate ID management systems for Dexpi Identifiers in each company, but with a coordination which makes re-use possible and easy across companies in DISC.
+
+
+That we are looking at P&IDs means we are identifying 
+"processing and instrumentations", so functions and activities 
+that usually have the same 
+lifespan as the facility, and not the actual physical equipment. 
+
+The identifiers are unique at least in DISC, probably globally, and cannot be re-used for example if the same P&ID is re-used in different locations.
+(An example of such re-use would be a P&ID of a component from a supplier.)
+
+
+### Requirements
+The ID system must have these capabilities: 
+1. Looking up an ID and getting basic information about it. (TODO: How much?)
+2. Creating a new ID together with basic information especially including all other relevant IDs, like tags.
+3. There must be a possibility for several ID management systems that do not overlap
+
+In addition, to make integration easier, it should support this capability:
+
+4. Checking if there exists an ID for a given "descriptor". F.ex.: "Is there an ID for the valve before tank P4712 in the Aasta Hansteen facility?"
+
+
+For example: 
+Equinor operates AHA and is planning to have a tank with Tag 4712. Aibel is designing a P&ID for the tank and wants to create an ID for the tank.
+Aibel get this ID from Equinors id system, which generates a new PersistentId: "Context equinor, Identifier 123"
+Two years later Aker BP takes over the operator role of AHA. Equinor sends over all IDs used on AHA and their metadata to Aker BP. 
+This means that "Context equinor, Identifier 123" is still available and usable, but can only be looked up in Aker BPs ID system.
+Aibel now is going to do a change to the tank they made for Equinor, and creates a new nozzle, which they asks Aker BP for: The
+query includes the "Context equinor, Identifier 123" for the tank, which returns metadata, f.ex. one or more P&IDs for 
+the area around the tank. The Aibel engineer sees there is no ID for the nozzle, 
+and asks to get a new Id "Context aker, Identifier 12345" for the nozzle.
+
+So the ID systems can contain and be used to lookup IDs that are generated in other systems, but can only generate IDs
+for their own dedicated context.
+
+This means there is no centralized global lookup, so you need to know the operator to get info about the ID. 
+This assumes that all users and systems are aware of what operator is currently operating each facility.(TODO: Is this assumption correct? )
+
+An alternative would be to route requests via the old operator to the new one, using a lookup table on the old one, or
+to encode facilities in the IDs such that a global router could be configured to route requests to the correct operator.
+
+An important question is what data can be stored in the ID system, since it probably can be used to reverse engineer the design.
+Can f.ex. all P&IDs be shared between all DISC partners? Or must this be operator specific and not DISC-global?
+
+### Using the PersistentID element from Proteus XML
+In Proteus XML we follow NOAKADEXPI and suggest using the PersistentID element with Identifier for the actual id,
+and Context for the name of the central ID management. 
+This name could be the URL of the web-accessible api of the system.
+