Friday, November 11, 2016

Huffman coders and routers

In the past, he two have been unrelated, mainly because route look-up tables were screaming demons, built of special hardware.  The congestion functions is simple, no congestion then you hit the machine limit and it blows up. They are like block chain, no compression, infinitely accurate.  As a descending graph, the number of 'route' in a bin, on any node, was simply set to what the screaming demon could handle.

Buy... when Cisco wants to automatically spawn a router, which is what they are attempting, then the shape of an ongoing probability look up table is critical, it tells you immediately how to segment the stream and drop a subgraph onto another machine.  It is also critical that their screaming demon look up macine be icrocoded onto therisk processor, make it run under the python interpreter.

One very key help is smart card, of course, the router system, with an adaptable Huffman look up table can spot a black bot, and identify it by pricing cycles on the look up table.  They get implicit compression, bots unwilling to pay the price time-out.

Their look up tables are just like the nested graph, which is just a standard method of making a graph optimized for rapid descent. They should get a good micro code of the look up, then adapt it to the Python interpreter.

They are a bit trapped by Moore's law, which has gone far enough to make micro coded risk machines a good enough substitute for screaming demons.  So there is a blur of layers, give mer a room of risk boxes, let me sort sqlite streams using nested key value, or let me sort IP streams using nested key value.  If both forms spawn equivalently to soak up congestion, hen  here is no layering.

None of this spooky

Adaptable look up table been around for ever, everyone knew hey were Huffman like.  It is simply that we have essentially intelligent 'packets' today, thank god.  So, self adapting queueing over networks is a possibility, no matter what you are routing, as long as the intelligent packets have a common, and enforcable view of cost on the graph. Then the look up table system gets a very good hint at the underlying congestion and can spawn the wotk load, that is, it can 'deal' with the packets, literally.

Ethereum becomes your network management company.  They enforce payments for cycles. If telecom cycles go as congestion, then the graph congestion should become coherent with telecom pricing. It is the Red/Grteen function that separates the routers from the stock traders.

No comments: