3 ms·
Some interesting parallels with your excellent comments on circles: The old Windows NT domain structure was based on one-way trust: Domain A could choose to tru
by e12e 11y ago
Some interesting parallels with your excellent comments on circles: The old Windows NT domain structure was based on one-way trust: Domain A could choose to trust domain B, while B might not trust A (along with a set of include/exclude rules which allowed for even more complexity). This is the kind of idea that might look like a great way to implement principle of least privilege. But it turns out, that it's just too hard to maintain - and even if you have an organization with such high demand for compartmentalization, you rarely need one-way trust; you can get most of the same/better result with isolated systems on one hand, and federated systems on the other (eg: you have a "red room" that isn't networked, with strict protocols of what goes in/out for sensitive stuff). Couple that with the fact that "one way trust" doesn't really work - if you can pull in something dangerous like a spreadsheet, you've open yourself up to dangers from the "untrusted" subsystem.
I think it is similar with circles: It doesn't make all that much sense for me to list someone as a "close friend" and they to have me in their "random co-worker" group. With network effects (people who are actual friends, talk together and show people stuff in the real world) - there's no way to really keep stuff that's shared with group A secret from group B - it might not give a false sense of security (hopefully most people by now realize that if something is shared, it's no longer secret), but I think it does give a false sense of control.
As to your point about salons - this is my experience with Facebook groups as well. FB is a terrible medium for discussion, but apart from the UI/UX - it's where people can be found, and moderated groups can and do work rather well. And they basically work like you describe salons.
- dredmorbius 11y agoMy argument to Google has been that independent controls for tagging content topic and controlling distribution are what's needed. In the real world our informational transactions are largely determined by place, and we assume different roles by those. Home, work, gym, mosque, store, school, concert (performer or audience?), beach, etc. And how and who you (or others) are known varies -- the idea of single unitary identity is often quite absent. Size of groups also matters hugely, and I think often of how dynamics change with scale. See the old Budapest String Quartet joke. Expertise, qualifications, and intent of participants also matters. Simply mashing bodies into a room, not so much.