[Chapter-delegates] hmmm...

[Narelle]

Fred, Bob

there are more issues here, than the principle of filtering.

1. Transparency - where ISPs make decisions to change the throughput
of traffic, and let's face it many, many do, then these should be open
and transparent to their users. There are many good citizen type
efforts, such as DOS attacks or other automated attacks,where it is
quite common for ISPs to filter addresses completely, reprioritising
traffic during outages, balancing of traffic for performance reasons.
Then there is the use of DPI prioritise for fee services, or outright
block competitor services.

The operational policies in place by ISPs are not often published, and
users/customers would not necessarily be aware that this type of thing
exists, nor that it happens. How does this type of thing stop? Are
addresses filtered forever? Or are they randomly removed, just as
randomly as it seems they are applied.

2. Network Dimensioning - where a single application can damage the
performance of other Internet users, there must also be the question
of whether the ISP has sufficient capacity to deliver the service it
is selling. If an ISP is marketing an 'unlimited' plan then surely it
is reasonable to expect that many of its customers will be heavy
users? Sure, in every network I've designed, I've always tried to
pre-empt it with a traffic model, and then ensure sufficient capacity
within the network once it is built through ongoing monitoring. One
doesn't always anticipate how a marketing department might represent
it.

We have had some success in Australia through the application of the
'Trade Practices Act' - where ISPs have been called to account for
making claims in their advertising that aren't the case in the
delivery. It still doesn't guarantee individual services perform on
any particular day... Bob is quite correct there.

3. Applications behaving in ways that weren't anticipated in the build
- eg continuous transmission. It is arguable that we didn't ever
anticipate applications that would consume bandwidth in the way that
they do. The fact is that networks are dimensioned taking into account
that you could apply sharing ratios up the chain - 1,000s users to one
uplink - if the majority of those users are continuously transmitting,
then the contention ratios change markedly. That might be okay if you
have many peering points and the bulk of the traffic is national, but
if that traffic then traverses highly priced international links, then
the costs  blow out.

Internet services have supposedly been too technical for users to
understand, and often that people have been keen just to get Internet
services, has meant that standards have potentially fallen by the
wayside. When arguably, network service providers have been able to
get away with being opaque for too long.

I suggest we discuss the nature of what Internet services mean. and
determine the sorts of principles that are fundamental to a genuine
Internet service.

I nominate transparency, any to any connectivity, and truth in advertising!

regards



Narelle

On Mon, Sep 27, 2010 at 2:39 PM, Bob Hinden <bob.hinden at gmail.com> wrote:
> On Sep 25, 2010, at 2:03 PM, Fred Baker wrote:
>
>> Net Neutrality question. I have heard some pretty strident statements on this list
>> to the effect that "ISPs should just let my bittorrent go and get out of my way". I
>> wonder what people think about this issue - what happens when the ISP lets
>> someone else's bittorrent go and gets out of their way, and as a result my service
>> is impacted? Does the ISP have any requirement to ensure that I get what I'm
>> paying for in my contract?
>
> As you know it's complicated.  However, my answer is that if a class of application
> is causing an ISP to not be able to provide service to it's customers, it is fine for it
> to provide some controls to slow down that class of application.  What is not OK is
> to slow down it's competitors application and leave it's own version of that
> application alone.
>
> I would also note that most customers do not have contracts with their ISPs that
> provide much of a guarantee of anything.  They say what a customer can not
> exceed, but they don't say what they can get.
>




-- 


Narelle
narellec at gmail.com