But really, the point this thread is making isn't about whether you approve of all of AT&T's business decisions, but rather whether their backbone engineers can properly manage a network.
I'm interested in the opinions of people who have managed default-free BGP peering before on whether AT&T does a good job, but I'm not so much interested in reading another dslreports thread about whether AT&T is a "good ISP".
There's always the chance they can try to put the customer on the hook for a truck roll, turning a loss into revenue.
As a service. The gateways are still crap. Don't use 'em for routing or wireless. They are laughably bad.
However, you throw a SIM card from your phone in the hotspot and you'll magically have an IPv6 address, even without changing the APN. It's ridiculous that I have to resort to using a VPN to bypass all the troubles being behind a CG NAT entails.
Some of the branch locations at work have U-Verse as their only option. I don't have to deal with their day-to-day problems but the help desk tells me they're still pretty crap.
Their wireless router and DSL modem "U-Verse" gateway device lacks a true bridge mode, has a small NAT table, and you must use it because it contains a 802.1x certificate. You can use your own router with a rather involved hack: http://blog.0xpebbles.org/Bypassing-At-t-U-verse-hardware-NA... They instructed customers (including small businesses) to get off 10.X.X.X, presumably because they were considering CG-NAT, although they have since done some IPv6 deployment.
The original AT&T faded away with declining long distance revenue which they milked to the end until cellphones and Internet access became the main market. It took tax credits meant for "video dial tone" broadband deployment and used them to build their cellular network.
Its ADSL deployment use PPPoE and ATM which can cause MTU mismatch. ATM itself is a telco mistake. European access oligopoly members wanted 32 bytes because that suited voice frame transport and they thought data transport would never sell. American counterparts wanted 64 bytes for better data delivery. They compromised on 53: 48 for payload and 5 for header. With newer "U-Verse" ASDL they've finally eliminated this layer of overhead. (But no more 3rd party modems).
It decided to deploy VDSL instead of make the true upgrade to fiber. But they also cut costs by not building enough remote terminals so many customers have slow and unreliable links. Fast VDSL speeds (e.g. 100Mbps) aren't possible if customers are a mile away, or more. How much money have they spent on truck rolls diagnosing issues? Is it less than just building a more robust network to begin with?
Since 2005 they've been complaining that customers are using too much Internet and they need to slice and dice the Internet with data caps, application specific throttling, etc. And get more tax credits and deregulation which they promise will be used for broadband investment.