[OSM-dev] 64-bit ID thoughts...
ebwolf at gmail.com
Tue May 24 06:17:34 BST 2011
I was just reading the recent thread about 64-bit IDs and had a couple
thoughts to put out there. 64 bits gives us a really big number space for
IDs. We should take advantage of this:
1. Keeping the current usage of negative IDs costs us only one bit. We are
currently only getting 31 bits. The jump to 64 bits actually increases the
number space by a factor of 4 Billion even while preserving negative IDs.
2. How about, as part of the transition to 64-bit IDs, we also shift to a
single unique object ID. That is, all changeset IDs would be unique from way
IDs which would be unique from node IDs, etc. One of the things the USGS and
US Federal Government has learned the hard way is that adopting a truly
unique object ID facilitates data synchronization. I'm not at all saying OSM
should adopt the crazy UIDs that the USGS is using. What I mean is just
allocate all IDs from the same 64-bit number space.
a. To make the transition, you'd probably want to "hash" the existing
object type into the ID. Add a "1" at the beginning for existing nodes, a
"2" for ways, etc.
b. I'm not a big fan of permanently hashing the object type into the ID,
so one sequence in the database would be adequate.
Just some thoughts...
Eric B. Wolf 720-334-7734
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the dev