 rchandraStargate Universe fanPremium
join:2000-11-09
14225-2105
1 edit | reply to Harddrive
Re: Complete Per Byte Billing. said by Harddrive:i want to see a low 'base price' for having the service available. then be charged if i use any on a per-gig basis. make it just like my electric or water bill. the less i use, the less i pay. but we know that will never happen. To some extent, this is merely an ideal. It's a great model, until real-world factors intrude.
It's kind of like socialism in this way: I think it's good conceptually, in that socialism helps out everyone, and it means those in unfortunate circumstances get some help from everyone else. However once real-world factors intrude, the whole thing begins to unravel, namely human nature. The incentive to produce declines as one observes that those who don't produce are taken care of anyway...so what's the point of producing when one can receive for doing essentially nothing?
Similarly, the real-world factor that intrudes on usage-based billing is, noone can control completely what data are delivered over one's link. If I decide I don't care what it costs me, I could (flood?) ping you, and even if you never reply to a single echo request on your end, you would still be paying for all the data I caused to go down your pipe. The only completely fair way to implement that is if you had control of every single Internet host, if you were able to "firewall" at the source. If you could even firewall on the other end of your link to your ISP, it wouldn't be fair to your ISP or any of their upstreams because they would be paying for the traffic being sent to them that you didn't want. It's a bit along the lines of what's described at lartc.org, about how one cannot control the rate of the receive end, only the rate at which one transmits. Yes, this indirectly affects reception because most higher level protocols (think "TCP") regulate bandwidth transmitted based on what's received (transmission is not supposed to go beyond the ACK window, but there is nothing absolutely preventing that).
To borrow another analogy at lartc.org, you can't control how much postal mail you receive. You can influence it by asking to be removed from mailing lists, you can never send out anything that would cause someone to reply, but you're not going to be able to stop every single mass mailer; there is nothing in the postal system to reject mail at the source. And once a mail piece is in the postal system, you can't refuse it at least until you've received it.
...which goes full circle to the point. In virtually all commerce, one plans for expected average demand. Based on trial and error, and previous experience, some quantity of "stuff" is produced, a lot of it (sometimes all of it) is sold, and many times there is stuff left over. In the case of a sellout, there is unsatisfied demand and potential profit is lost. In the case of overproduction, either usage of space to store the overproduction for later sale causes loss in profit, or if the stuff is perishable, it is simply destroyed. For example, the power company you mentioned doesn't know ahead of time precisely how much electricity you will consume in a given month, but it has a pretty good rough estimate, so it will go and purchase approximately whatever oil, coal, nuclear fuel rods, or whatever to meet the demand they expect from you...and it won't always be right, and maybe on rare occasions not even close.
So too I have to believe it is for ISPs. They estimate that for some forecast period (say, the next year) they will have to carriage X bytes from point A to point B, which will require Y in electrical purchase, Z fibers to be laid, XX in real estate to lay that fiber across, YY in personnel to set up and maintain the equipment, and so on. They don't go laying another fiber from me to Canonical so I can download a copy of an Ubunutu DVD this month, then take it up after I'm done. Nor would they want that fine-grained procedure to satisfy that demand; it'd be too expensive. No, they provide some reasonable capacity that I pay some fairly reasonable fixed fee per month for, and the world is mostly happy. If I want more capacity (as in, greater amount of octets carried to/from me per unit of time), I can ask for and pay for that additional capacity.
It just so happens that that we don't typically pay for and measure in the same intervals; we tend to measure capacity by the second but only bill and pay for by the month. Still, in most other industries, they manage to survive and even thrive on the same estimate, produce, sell, repeat cycle. I personally don't see why ISPs can't continue to do that.
Edit: One thing that ought to be considered very carefully is that there will be increased costs in installing and maintaining metering systems. Today it is very simple: in my case $41.95 goes from me to TWC every month, in exchange for a 10/1 M/s connection (and no caps or other technical limitations as far as I know). Anything more complex than that will require expenditure to implement, and I don't quite see how it's so necessary (in essence to change the operational and economic model). |
