[malandrew]

We need forward thinking investors like pmarca to drop some of the VC cash on the W3C standards process. We have the most amount of innovations and investible companies when there are more open APIs. The image tag, the current implementation of which was first coded by pmarca, enabled tons of startups, the biggest of which are companies instagram, facebook, imgur, etc.

The one API that is languishing at the W3C that would help us immensely is the Contacts API. Right now your contacts is effectively owned by companies like Google (via Android), Facebook, and Apple (via iOS). Every other company that wants access to your address book for the purpose of a social experience, needs to go through those three gatekeepers 99% of the time.

The ideal gatekeeper of your contacts should be the browser. The term for them is "user agent", i.e. it acts as an agent on behalf of the user. Right now there are a lot of "skills" these user agents lack. Investing in giving them more skills creates more decentralization and debases the power of the giants we resent.

[roberthahn]

I think you're right about using the web as an interop protocol. I want to see that future (or that present, more evenly distributed).

But I believe the browser is NOT the ideal gatekeeper. The ideal gatekeeper for your contacts is, surprisingly enough, an app designed to manage contacts. And as long as the inerop protocols are respected, switching from one contacts app to another will never be a problem.

[malandrew]

Browsers must be the gatekeeper, otherwise Facebook continues to be the only way to get your contacts into web apps.

What browsers need to do is leave more of the experience to end developers and expose as many low level APIs in a safe way. Browsers would do best if they focused on a sane approach to ACLs.

Browser plugins like SafeScript for example one basic way in which things could be better for users. What SafeScript lacks is reputation information on resources to help non-technical users make decisions about what to trust and what not to trust.

e.g. Alice and Bob are friends. Carol is a tech professional with a stellar reputation. Alice is tech savvy. Bob is a luddite. When Bob is presented with an ACL request for an unrecognized resource (such as an app or script from an unrecognized domain), Bob should be able to check if either Alice or Carol decided to trust that script.

Reputation systems, the web of trust, organizations like Spamhaus, EFF, Mozilla etc. can all go a long way to helping users make sense of what they can and cannot trust on the internet.

The ideal user agent would be like a docker container with an ACL for taking sensitive user information and sharing it with whatever is running in the container in a safe sane way that puts the user's safety and experience first.

I would love to see someone take the following things/features and mash them up:

* docker/lxc * chromeless browser windows controllable via API and any programming language (not just javascript) * QT like windowing system with URI routing and skinnable with the good parts of CSS. * ACL * reputation system for resources with URIs * Incrementally loadable

Linux containers provide the ideal technology to reimagine what the web could have been if Kay [0] and Engalls vision had become the predominant way of internetworked sharing of stuff.

[0] http://www.drdobbs.com/architecture-and-design/interview-wit...