Axel Boldt <[EMAIL PROTECTED]> writes: > I would like to advocate for a multithreading/parallel download > feature and I believe the last quoted sentence above is simply > false; parallel downloading provides considerable speedups in almost > all settings.
The most noticable speedup in the single-server case is when you share the bandwidth with other persons or programs. In that case, if a single download is using 1/n of the available bandwidth, two will use 2/(n+1), which is obviously more. However, this approach is somewhat questionable because it is a way of "cheating" -- on a shared connection, you get the additional speed by making other people's downloads that much slower. If everyone were using such mechanisms, you'd end up back to 1/n, with downloads being slower for everyone to boot (a tragedy-of-the-commons-type situation). > All this is on single processor machines. I don't think processor speed plays a role here. In Wget's case, the data still arrives from the Internet much slower than the processor is able to process it.