> On 2 Nov 2016, at 19:56, Esteban A. Maringolo <emaring...@gmail.com> wrote: > > 2016-11-02 15:48 GMT-03:00 Offray Vladimir Luna Cárdenas > <offray.l...@mutabit.com>: >> Hi Sven, >> >> In our ongoing Data Week, we found that would be useful to use a unique >> identifier for objects that is not integer (some kind of SHA or similar). >> The use case was this: >> >> We were editing the same Grafoscopio tree in a DVCS (fossil, but could be >> git or something similar) and the nodes changed in a conflicting edition. >> Contents were different, but the order was the same. Do you think is useful >> this kind of unique ID index to make STON even DVCS friendlier? > > The ID for references should be some form of UUID/GUID. > SHA or any other hash wouldn't work since it would be dependant on the > contents of the object. > > However this would increase the file size considerably, 38 bytes per > id (32 hexadecimal leters, 4 hyphens and the quotes for the whole > string). > > However according to what Sven said, the integer used for the ref is > the "order" of the element, so it would be a whole different > referencing strategy.
Right (the last conclusion). The reference identifiers in STON are only used when there are actual references needed (in the case of shared instances or circular references). Furthermore they only make sense in one particular graph. If you want some form of external UUID you should add them to your objects. > Esteban A. Maringolo >
