Bruce Miller wrote:

We have had thoughts about ADSL. At present we are still experimenting with Smoothwall.

Ugh. Smoothwall has had a bad taste since its project lead went beserk at me. I tend towards IPcop for standalone and shorewall on machines with full distributions on, myself.

Trialling 'smoothy' on a Rockwell serial modem has been a pain though. In fact modems and Linux seem to be a pain altogether, I have to say. (Polite suggestions welcome on this BTW).

There's still no substitute for getting the correct settings from your ISP. That said, wvdial has worked fairly well for me when people have been unsure. MJR

Phil Driscoll wrote:

On Monday 05 May 2003 10:25 pm, MJ Ray wrote:

...

Ugh. Smoothwall has had a bad taste since its project lead went beserk at me. =2E..nevertheless, Smoothwall is a very good product, and certainly gets =

A product that has used the \d*.\d*.\d*.\d* regexp or something similar to check for a valid IP address (hint: it doesn't) and has installed hard-to-remove nag screens in the past isn't a good one, is it? It's probably improved, but other things are starting from ahead of it and are easier to get help with. For most people, I agree that a router is easier, but which routers make it easy to NAT and run services from inside? -- MJR http://mjr.towers.org.uk/ IM: slef@jabber.at This is my home web site. This for Jabber Messaging. How's my writing? Let me know via any of my contact details.

On Tue, May 06, 2003 at 04:20:42AM +0100, David Bowles wrote:

...err, I didn't realise this. Is this specification practice documented anywhere on the web?

...

Contention are frequently misunderstood. 5:1 actually means something more along the lines of 250 2M connections being multiplex over an ATM network onto one 100M ethernet port at the ISP end.

So I assume one also needs to take into account the maximum real-world

Why, since the figures arn't exactly based on "real world" throughput, in the first place? They are theoretical "worst case senarios".

throughput of a 100Mb ethernet port is nothing like 100Mbits / second. Most likely this is going to peak at 20Mbits / second, max. What this implies is when applying this 'hypothetical' definition of contention ratios, in practice this is more likely to become 25:1.

So far as the calculations are concerned the figure used is more likely to be 100M than 20M. Or even 45M to a router with an ethernet interface. (In which case the ADSL side would be up to 112 2M, 225 1M, 450 0.5M or indeed any mixture of these totalling up to 225M.) The basic point is that you don't have anything like 5 2M ADSL lines connected to one 2M line at an ISP.

Also, if the ISP decides to simply slot a 1Gbit ethernet link into the data stream, can they legitimately claim their users now benefit from an uprated link with a contention ratio of "0.5 to 1" ?!!!!!!!

No, since that wouldn't change the rate of their connection to the ATM network. If they did change that connection from 100M to 1G BT would feel free to direct 10 times as much traffic to that port. The way it works is that the ISP has one or more connections to BT's ATM network. N.B. the ISP connections have considerably greater bandwidth than an ADSL line. A connection ratio of 5:1 means that BT can direct ADSL lines to a total bandwidth of 5 times the bandwidth of the ISP connection to that connection. This figure won't tell you anything about contention ratios for the ISP's peering bandwidth or contention ratios for a specific ISP service. e.g. a filtering proxy. Which are probably far more relevent than anything to do with the ADSL figures. Especially with a general purpose ISP which has a broad range of different types of customers. -- Mark Evans St. Peter's CofE High School Phone: +44 1392 204764 X109 Fax: +44 1392 204763