[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: shim-aware transports
On 17-aug-2005, at 17:31, Spencer Dawkins wrote:
when we were discussing TCP in TRIGTRAN, we pretty quickly
converged on NOT notifying transports when bandwidth available
changed, because TCPs have to respond to losses, and adding
additional hints just adds stack complexity beyond the complexity
that is unavoidable.
Well, I'm sorry to have to say it, but that's just STUPID.
What if I have a 119 MB/s gigabit ethernet link and a 1 kB/s GSM data
connection, and I'm downloading a nice big file with my fully RFC
1323-compliant IP stack, and then the gigabit link goes down so I
rehome to the GSM link.
Suppose the file is coming from 100 ms away, so my window is 12 MB or
so. That means that worst case, my GSM link could be saturated for
more than THREE HOURS because of that single window worth of data
that's in flight.
This "complexity is scary" belief in IETF is getting ridiculous and
then some. (While at the same time you need to read some 20 RFCs to
implement SNMPv3. Good thing they never went for a "complex network