On Fri, Mar 28, 2014 at 9:56 AM, Stephan Beal <sgb...@googlemail.com> wrote:
> On Fri, Mar 28, 2014 at 2:24 PM, Richard Hipp <d...@sqlite.org> wrote: > > > SQLite prefers small positive rowids, since such rowids take up less > space > > on disk using the variable-length integer encoding. ( > > http://www.sqlite.org/fileformat2.html#varint). But there are no > > guarantees. > > > > i guess we can fix Fossil's assertion's in about ten thousand years when > the record IDs roll over into negatives ;). > The record IDs will overflow a 32-bit signed integer long, long before they overflow the 64-bit rowid. The current "burn rate" on the SQLite repository is about 3650 record IDs per year. Let's assume 10x the burn rate. Even then, it is another 58,000 years or so before the 32-bit signed integer overflows. At that burn rate it is 2.5e14 years before the 64-bit rowids run out and need to go negative, which is about 18,000 times the estimated current age of the universe. By the time the record IDs overflow 32-bit integers, the repository size will be over 2 terabytes. -- D. Richard Hipp d...@sqlite.org _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
