Hi, On Wed, Feb 6, 2013 at 2:25 PM, Michael Marth <[email protected]> wrote: > Excuse my ignorance, but I never came across the concept of a tree of journals > before. Is this an entirely new idea or has this been around somewhere?
It hasn't come up in exactly the same format before. The inspiration comes from generalizing the branch concept, the write concern feature in MongoDB, and some discussions I had especially with Thomas last year. > One other question I had: how does this proposal relate to the proposal for a > ID-based MK API [1]. I realize that they are generally orthogonal to each > other (one being about API, the other about implementation), but I wondered > if the segment-based implementation might benefit from ID-based access. > Any thoughts on that? Yes. In the proposed design each node state can (and should) be directly addressed with a "<segment UUID>:<offset>" record identifier without having to traverse its path from the root. BR, Jukka Zitting
