Compression question / problem

[Dave Cragg]

On 12 Jan 2005, at 23:03, Alex Tweedly wrote:

> Dave Cragg wrote:
>
>> Unless the files are on different servers, It won't be any faster to 
>> download in parallel using "load url". For requests to the same 
>> server, "load url" queues up requests, and when one completes, it 
>> starts the next one.
>
> Ouch !!
> I didn't know that (and assumed the opposite). Thanks Dave.

>
> Does "server" mean name or IP address ?
> Or even "IP address+port number" ?

It's numerical IP Address + port number.

I'm not sure if "thanks" are appropriate. Perhaps libUrl could be 
improved in this respect. However, I'm not sure if there would be any 
great benefits from having parallel downloads from the same server. My 
understanding of pipelined http requests (i.e. simultaneous requests 
over the same connection) is that the data must be returned 
sequentially in the same order as the requests were made. To get truly 
parallel downloads, separate connections (sockets) would be needed for 
each request, and that has its own overhead.

Right now, libUrl uses  a "conceptually simple" approach (i.e. the most 
my brain can handle) while still taking advanatge of re-using 
connections (in the spirit of the http 1.1 spec).

Cheers
Dave