Hacker News new | past | comments | ask | show | jobs | submit login

> but rather for the OS to be aware of when it's connected via a metered/capped connection, and behave differently.

Pretty much all internet connections are metered/capped, though prior to recent FCC transparency rules the caps on residential fixed broadband were often undisclosed or affirmatively misrepresented under false "unlimited" labels.

Making the OS appropriately sensitive to the costs incurred by each network-involved transaction relating to updates and the system owner's preferences regarding balancing those costs against the value they provide in update experience is abstractly ideal but decidedly difficult.




As i was saying above, the majority use-case of update sharing is sharing updates over a LAN to other computers in the same office.

Note that, even in that case, the connection to the outside world might still be metered/capped—but the connection to LAN peers obviously isn't. That means that "does this traffic cost anything" is an evaluation the OS would have to make per socket (or requested socket from a higher-level library, like Windows' BITS), rather than per interface.


The point you are making about per-socket evaluation here is what I was trying to get at with the "network-involved transaction" reference in GP.




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: