10 ms·
Sounds like a bit of an unpopular opinion in this thread but I think pay-for-what-you-use plans with reasonable per GB charge and a slightly higher per GB charg
by kinkrtyavimoodh 6y ago
Sounds like a bit of an unpopular opinion in this thread but I think pay-for-what-you-use plans with reasonable per GB charge and a slightly higher per GB charge above a certain number seems reasonable, irrespective of your opinion of Comcast.
Something like 30 dollars for a residential connection, followed by 2.5 cents per GB for the first 1024 GB, and then increased to say 4 cents per GB for another 1024 GB.
If you use 1 TB, you pay 30+0.025*1024 ~ 55 dollars.
Why should we subsidize those who download tens or hundreds of TBs?
No one thinks this is unreasonable for electricity, why is it unreasonable for data? Bandwidth issues affect data too.
- charleslmunger 6y agoYou might find this comment thread, which discusses that same analogy, interesting: https://news.ycombinator.com/item?id=17426607 https://news.ycombinator.com/item?id=17426607
- gruez 6y agoThis comment was posted two minutes ago and it's already showing up as grey to me. It's a shame, because what was said has merit, even though the conclusion might not be appealing to the average HN reader. eg. >Why should we subsidize those who download tens or hundreds of TBs? Consumer internet is oversubscribed. By using 10x as much bandwidth as everyone else, you are effectively getting your internet connection subsidized by everyone else.
- RileyJames 6y agoA reasonably high cap like this doesn’t seem completely unreasonable. But where the cap is set, who it applies to, when it’s updated and when it comes into effect are more important than the number of TBs. For instance, setting it at 1.2 TB today seems fine, but in 10 years that won’t be nearly enough, but will Comcast update it? Or just push ‘heavy’ user to their ‘prosumer’ plans. It’s similar in financial regulation. Transactions over $10,000 must be reported for KYC/AML purposes. Sounds fine when it was implemented, but now it seems very low. It wouldn’t take much to future proof decisions like this, by default. Inflate by 10% per year, or some other metric. And the lack of such features hints at the fixed amount being the feature.
- fomine3 6y agoI'm interested in UMA in mexico. https://www.littler.com/publication-press/publication/mexico-increase-uma-value-announced-2020 https://www.littler.com/publication-press/publication/mexico...
- toast0 6y agoHigher prices per kWh after some limit is a regulatory tool to drive down energy use. There's no regulatory reason to drive down data use. Given that, the 1st data allotment should include the cost of the last mile connection, and any subsequent data allotments should be significantly lower cost, closer to incremental transit costs (which aren't very much). Commercial internet connections are often billed at a monthly fee based on the interconnect type, and then a per Mbps fee for 95 percentile usage. Most residential customers don't really want that though, they just want an unmetered connection that works, with a monthly price they can count on; and one where the more they pay per month, the faster it goes, and the more stuff they can do. A flat cap set across all of the tiers doesn't make sense. And how are consumers expected to control all of their devices to make sure they don't download hundreds of gigs of updates overnight?
- rasz 6y agoSubsidi-what??? You do realize transfer in bulk is peanuts? $55/month is more than whole dedicated i7/SSD server with unmetered 1GBit at Hetzner.
- Const-me 6y ago> Why should we subsidize those who download tens or hundreds of TBs? Why should those who download tens or hundreds of TBs subsidize the rest of the people? End-user equipment, underground cables, and ISP's routers ports are the same regardless on the bandwidth. Construction work is expensive, you need American people to lay cables in American soil, can't outsource to India. Compared to these costs, bandwidth is borderline free, even in the long run.
- jjk166 6y agoYour internet plan is already pay for what you use: you pay for a given rate of data transfer and your ISP maintains the infrastructure to provide you that rate. The cable you need for 50 Mb/s costs exactly the same whether you use it 2 hours per day or 20. Anyone downloading 100 TB in a month is already paying a hefty premium for the (at bare minimum) 370 Mb/s they need to achieve that download. For electricity, in part you are paying for the infrastructure, but you are also paying for the fuel that's being consumed by the powerplant. If comcast charged per GB of data that it hosted, that would be fine (and hosting providers do exactly that without complaint), but Comcast's costs don't vary with the amount you download. In the extreme, costs do scale to some degree with duty cycle - a line with lower capacity can service more people so long as they don't use it at the same time, and technically sending a packet through causes some infinitesimally tiny wear and tear. In practice though, a data cap doesn't stop multiple people from doing their 1.2 TB download all at the same time, and the reduced lifetime of components is a rounding error compared to the variability just from their manufacture.
- em-bee 6y agoi had a pay per use plan more than a decade ago. i liked it because the competition all had capped plans that cost the same as my pay-per use would if i reached that cap. and if i didn't, then i paid less. the service also allowed me to set my own cap so that i would not accidentally go over a chosen spending limit (and it allowed me to change the cap at any time)