Richard Rauch <rkr@olib.org> wrote:

> Hi.
>

Hi Richard, 

>    Did you repeat the test, to verify that it was not a statistical
>    anomaly?

I repeated the test again tonight.  Again, a small 
improvement showed with the larger buffer.  

>    I still divide by 10 to figure the max. throughput.  So if you ever
>    see close to .5 MB/sec, I'd break out the champaigne and call it a
>    red-letter day.  (^&

I'll keep a bottle available.  :-)

>  * The idea about using multiple connections is a good way to test
>    your real performance.  

David's idea.  :-)

>    But multiple connections to a SINGLE HOST are
>    more likely to be throttled (by that host).

Hmm.  It seems like a lot of extra work for a server 
always to be keeping track of the sources of all the 
connections and ensuring that each connection is from 
a unique client.  But I'm pretty ignorant about servers, 
so my opinion doesn't count.  

>    You should probably do multiple connections to widely scattered hosts.
>    I suspect that you will find that you can approach .5MByte/sec if you
>    do that.

We'll see.    

> My personal opinion is that 5Mbits/sec is far enough ahead of the
> curve for general Internet users (ignoring large businesses, educational
> institutions, etc.) that you will find it hard to see that kind of
> performance in practice.

I think, if the ISP sells "up to 3 Mbps" and also "up to 
5 Mbps, then it's reasonable for the customer to assume 
that she should be able to download at something approaching 
the rated speed at least occasionally.  Otherwise, they 
could just as well sell the connection as "up to 100 Mbps" 
or "up to 1000 Mbps."  

I must stop ranting about this!  :-)  Last time, I promise!  

> If you're not happy with the performance that you're getting, you
> might consider returning to 3Mbits/sec and waiting a while
> (say a year?) while the network infrastructure slowly improves.

Now there's a good idea.  

> Good luck.

Same to you, Richard.  

Kindest regards, 

Tom