On Fri, Sep 12, 2003 at 01:37:59AM +0200, [EMAIL PROTECTED] wrote:
> >The difference is too small to be meaningful.
> when you have 56k modem and ~4.5KB throughput 0.5KB can make a big 
> difference

I wonder if we're thinking of the same numbers. I've seen 56k modems 
run at 14,000 bps on bad lines, but not often. 4,500 bps is pretty 
> >The change means that the second connection traveled over wires that were 
> >slightly noisier, or slightly busier, or one of a hundred other 
> >possibilities. The odds that there is a difference in connection speeds 
> >caused by a difference between ppp and pppd lie somewhere between zero and 
> >null.
> I tried it several times and got the same results, therefore it must be 
> caused by the diffrence between ppp and pppd.

I've done repeated measurements on connections, gotten consistent results, 
and discovered a week later that I was getting different results. I would 
not trust results unless they were consistent over time. I don't 
know how many tests you did or over how long a period you did them, so 
you may be right. You didn't mention repeating your test in your first 
post, and I'll admit I assumed that you did it once. 

Running ppp in the kernal is supposed to be more efficient. In practice, 
I doubt it makes a lot of difference. Most of what controls transmission 
speed is in the modem, not the computer. Some things that affect speed 
can be negotiated between the computers, and that might make a difference, 
depending on how you're configured. Or I could be just plain 
wrong; user PPP might have been tweaked to the point where it works 

I used pppd until a few months ago, and my recommendation is to go 
with user PPP, regardless of speed differences. Few people use pppd, so 
you won't get as much help if you have problems. Roughly 75% of the 
answers I got when I was asking for pppd help explained how to fix the 
problem if I were using user PPP. User ppp is much more flexible, and I 
did a lot of work solving problems on a couple of occasions that could 
have been solved quickly and easily if I had been using user PPP.

Bob Hall 
[EMAIL PROTECTED] mailing list
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Reply via email to