Seeking feedback on proposed D-Base32 encoding
Written for Novacut by Jason Gerard DeRose on 2013-01-22
We're seeking feedback on our proposed D-Base32 encoding.
The problem with standard RFC-3548 base32 encoding is the sort-order of the base32-encoded IDs is different than the sort order of the binary IDs.
Eventually we're going to have Dmedia-lite... a trimmed down version aimed at mobile devices, and more at content consumption than content creation. For space and performance reasons, Dmedia-lite will probably decode the base32 IDs and build its indexes according to the compact, binary IDs. In fact, even the full version of Dmedia will probably do this eventually, because it will give us better performance and allow us to handle even larger databases.
Still, indexing according to the base32-encoded IDs is simple and useful, something we want to support. But there are serious interoperability problems between a database using binary IDs vs one using base32 IDs, because they will have a different idea of what the first and final document IDs are in the database. So a request for a range of docs like:
1 <= doc_N <= 6
Would come out nonsense on the other end:
14 <= doc_N <= 3
Click the "Read more" link for all the gory details.
Updated . Read more
