Wayne Bouchard [SMTP:firstname.lastname@example.org] writes:
The complaint ratio between the two groups are gonna be
wholly lopsided. The smaller ISP will receive far more complaints than
the larger one.
I disagree; subscribers of smaller ISP dedicated services expect a lower quality of service than that of a NSP like UUNET.
As a subscriber of both (UUNET T1 being turned up this weekend, and Probe Internet, a local ISP which has a T1 to MCI), I am more tolerant of routing issues, performance, etc. with Probe (who I pay approximately 20% of what UUNET charges) than I am with UUNET. Noting that I have been quite pleased with UUNET's service so far and generally do not have any service issues outstanding with them, I am concerned about the quality of service in relation to the alleged settlement scenerio discussed in this thread.
In reviewing the service literature from UUNET, I can find no reference to "partial Internet service," "not really the whole Internet, but rather what we could offer today based on our peering arrangements, which may vary from day to day, and we are under no obligation to tell our customers what we can and cannot offer," etc.. I can't find the web server on UUNET that provides bulletins about what networks are unavailable today, nor does there appear to be a major-domo list, so when my customers complain about not connecting to our network from anywhere in North America (through various ISPs), I can verify that their ISP hasn't paid UUNET for peering, etc..
I view such partial service as unacceptable from an NSP, especially when I can obtain full service from other NSPs for the same or less money. Unless I'm behind the times, MCI, BBN, PSI and others still have rather complete networks.
More fundamentally, I am concerned that NSP transit providers are loosing focus of the origin of their "peering value." If all UUNET, or any other provider for that matter, had to offer was their internal network (i.e. www.uu.net), I don't think many peers would be demanding access to their AS. (They've got a nice web site, but it's no yahoo / microsoft / etc.) The true value of their network is in the application services offered by their clients. Internet consumers demand access to yahoo.com, infoseek.com, msn.com,carpoint.com, etc., not AS284, AS1240, etc. Transit is just that - a pipe.
If transit providers start realizing revenue from providing transit to my application services, and in the process also create black holes associated with the respective policy, I can quickly envision a settlement model enforced by application service providers. Such models would likely be created to allow us to "patch" the holes, but would quickly migrate to revenue-producing models.
For example, as our company extend our FurnitureLink and Intellibuy satellite network into the Internet, delivering our product to retailers and consumers, I can envision a scenerio where we permit connectivity with transit providers and charge for packets consumed by each AS. Perhaps AT&T provides us with a DS3 and qualifies for a 12% discount on the packet rate; they can in turn offer discounted transit to peering ISPs. Our network may be substantially cheaper
for an ISP routing through AT&T than it will be for Sprint or PSI.
Anyone care for tremendously complex routing policy based driven by application provider settlements? Given that UUNET is an entry in my expense column today, and my service may provide them with additional settlement revenues, I would expect them (and any other transit provider that benefits from our application services) to move to the other side of my balance sheet.
Any thoughts on this? Explanations as to why I should pay UUNET or any other NSP so they can sell connections / peering to my application services? Perhaps Yahoo, Infoseek, MSN and such should charge their NSPs to interconnect? Is this already being done? I recall that MTV was evaluating implementing transit charges last November / December. Perhaps a RA for application service providers is the next step? Any others game?