Can I reserve / allocate official osm ids for private use?

Sorry, but AFAIK there is no way to reserver/lock IDs without real payload.

If you start from an empty dataset, you might use your own (nonnegative) IDs and keep them in a local xml file. Another way would be to setup your own rails port server with own API etc. that keeps your stuff seperated from OSM DB, while allowing multiple users to edit this data.

I'm not an expert but I guess there might be legal problems if you mix OSM data with non free (as you aren't allowed to publish) informations on a dataset level.

There is also a small but nice existing indoor community at OSM :)

Use positive IDs for your own data file, starting at 1. If you need to mix your data with data from OSM for some processing, use a small script to re-number your file to using IDs above the highest in OSM at that time.

Reserving IDs is a very bad idea... Aside from the obvious management complications, there are other complications whenever you encounter conflicts of IDs, software not aware of the reservations, etc.

Also, you're trying to use a tool (JOSM) that wasn't designed to work with "mixed" (OSM and private) data. JOSM will keep renumbering your IDs internally, and it wouldn't matter to it, because even if different between different loads of the same file, it'll be consistent at the moment of the upload to the OSM servers (which will replace the negative numbers with valid IDs)

The ID that'll be used is the next sequence number of the relevant tables. If by some reason the upload fails, the whole transaction is rolled back and the IDs used are "lost" forever.

However, not everything is lost... One workaround I thought is relatively simple, emulates the reservation system, but has some limitations. Let me explain...

• Change the sequences in your local copy of the OSM database, setting the next value to a very high number. The IDs are all bigints, use something like 1 x 10^18, totally out of reach of anything that "normal" OSM will reach in the next billion years. And still leaving trillions of IDs for your usage.
• Install the full rails port, including the APIDB server, etc.
• Modify JOSM to point to your LOCAL server. Very easy, in the configurations. Maybe, even change the look and feel (also in the configurations) to make clear that data processed with this version should NEVER be uploaded to the public servers.
• Add your data and upload to your LOCAL server.
• Update using the daily/hourly/minutely from the official OSM.

The limitations are:

• Your data will never be compatible again with the OSM data without more processing - the IDs would have to be stripped and replaced with negative IDs before it could be uploaded to the public OSM server. And probably there'll be many conflicts. Maybe someone else will do a survey and upload some version of a map to the same place?
• Your data will have to be fully new. Not even one node, one way can be shared with the public OSM database, not because of copyright, but because if someone changes this item in the public OSM database, your private data will be changed as well, probably in ways that you do not intend to.
• If someone else uploads something in the area you privately mapped, it'll overlap your data, and you can't remove it without almost surely breaking the possibility of further updates to you LOCAL database, because it'll fail when it doesn't find the conflicting items.

Updates to the APIDB using osmosis, or other tools, from the daily/hourly/minutely changes all come with the IDs already assigned, therefore the sequence number is never touched when you apply them. Your own changes, on the other hand, do not have valid IDs yet, and the API server will use the default (implicit or explicit, can't remember) of the ID column which is "nextval([name_of_sequence])".

EDIT: I didn't fully understood Frederick's answer cited before. Corrected!

Why do you even worry about conflicting IDs if you don't intend to upload your results anyway ? Put your objects in your own db, and use a mashup to display your indoor map on top of the osm basemap.

You can "seed" your work using an extract of OSM data if you want, say the town your school resides in. That would just be a snapshot of osm data; updating said data after you modified it should be possible, but is likely to run into problems now and then. Also, remember that if you seed your own map this way, it certainly constitute a derived work and you would have to redistribute it, making the whole effort of keeping your work separate a bit pointless.

If you ever want to merge your data with the osm data (assuming you didn't seed it with osm data, and kept it separate), just extract your db as a *.osm file, negate all the ids, load it into josm, check for problems, and upload to osm.