3 ms·

But what about the "network effect", that necessarily the thing that gets chosen to be scaled up cannot be optimal. TCP is not optimal (there are better protoc

by platz 13y ago

But what about the "network effect", that necessarily the thing that gets chosen to be scaled up cannot be optimal. TCP is not optimal (there are better protocols), bitcoin is not optional (there are better altcoins), but it's better to build layers of abstraction on this base (i.e. http) than to try to replace them.

andrewflnr 13y ago
Where do you get, from the network effect, that "necessarily the thing that gets chosen to be scaled up cannot be optimal"? The network effect just means that whatever is currently scaled up is hard to displace.
- derefr 13y ago
  In fields where we gradually refine our definition of optimality, we don't tend to be aware of the retrospectively-optimal solution until after we've put previously-considered-optimal solutions through their paces, and found their pros and cons. The network effect, however, tends to kick in when a solution is "good enough," which happens on some random iteration possibly long before the "truly optimal" solution is found. The only time you'll see an optimal solution to a problem spreading by the network effect, is when the definition of optimality is clear-enough from the start that the optimal solution can be found before people have a chance to settle on anything lesser.
agumonkey 13y ago
What other 'better than TCP' protocols have been tested in very large scales ? Just curious I never hear of any real competition for TCP. (and the few times companies, say Google, attempts to upgrade one, it doesn't end up that much better)
- derefr 13y ago
  When people say this, they're usually talking about SCTP. It's used by mobile carriers (especially as a control channel in 4G signalling), banks and stock exchanges, and in WebRTC (both as plain SCTP for the data channel, and in the form of RSTP for the media channels.) Google's SPDY (and therefore HTTP2) is basically an attempt to get the properties and advantages of SCTP, but only for HTTP, and done on a slightly higher level of abstraction where it isn't as efficient.
  - agumonkey 13y ago
    Thanks, first time I see this acronym and its influence on SPDY.

by platz 13y ago

andrewflnr 13y ago

Where do you get, from the network effect, that "necessarily the thing that gets chosen to be scaled up cannot be optimal"? The network effect just means that whatever is currently scaled up is hard to displace.

derefr 13y ago
In fields where we gradually refine our definition of optimality, we don't tend to be aware of the retrospectively-optimal solution until after we've put previously-considered-optimal solutions through their paces, and found their pros and cons. The network effect, however, tends to kick in when a solution is "good enough," which happens on some random iteration possibly long before the "truly optimal" solution is found. The only time you'll see an optimal solution to a problem spreading by the network effect, is when the definition of optimality is clear-enough from the start that the optimal solution can be found before people have a chance to settle on anything lesser.

agumonkey 13y ago

What other 'better than TCP' protocols have been tested in very large scales ? Just curious I never hear of any real competition for TCP. (and the few times companies, say Google, attempts to upgrade one, it doesn't end up that much better)

derefr 13y ago
When people say this, they're usually talking about SCTP. It's used by mobile carriers (especially as a control channel in 4G signalling), banks and stock exchanges, and in WebRTC (both as plain SCTP for the data channel, and in the form of RSTP for the media channels.) Google's SPDY (and therefore HTTP2) is basically an attempt to get the properties and advantages of SCTP, but only for HTTP, and done on a slightly higher level of abstraction where it isn't as efficient.
- agumonkey 13y ago
  Thanks, first time I see this acronym and its influence on SPDY.