michaelsevilla / malacology-popper

2 stars 0 forks source link

paper: introduction needs to tie graph to overall message #78

Closed michaelsevilla closed 7 years ago

michaelsevilla commented 7 years ago

Introduction: this section should show that programmers are trying to achieve programmability but it’s a hack -- the solution is expose some of the guts of the underyling storage, with a style for doing this in a principled way -- something that programmers are already doing (.e.g, using objects as an escape hatch) Peter: we can’t assume that these new object interface are for performance I’m not really sure where ceph ends and malacology ends. Lets address a concern of the reviewrs. Programmabilty is something users want and allows users to take advanatege of app specific semantics to increase performance and other properties like correctness. Expose guts to enable programmability in a way that doesn’t expose correctness as a concern Our contribution: What are the basic building blocks and where do we draw the line to enable awide variety of programs Do we have any a way to say that what we program in an object is orthogonal to correctness The sequencer is needed for corfu, but its a building block taht can be used in a lot of different contexts. Its less interesting if we built it for two use cases, but guess what, everyone needs a strongly consistent directory service. Decoule introduction from Ceph and talk in general terms. Principals we are proposing. In a separate section show what the evidence is. TODO: add finer granularity in how the object classes are being used. More detail than just categorical.