 |  | SIGNAL POSTED BY carl | 07.09.02 @ 04:46 PM
FAQ.3: Gosh, those expenses are really high. Is this some kind of gold-plated, cost-is-no-object deal?
[ FILED UNDER: Answers » Public Works » ] |
Er, we prefer "stainless steel" instead of "gold-plated." We're very careful with money (check out our audited financials and 9 years of tax returns). And, unlike the other bids, none of the .org money will ever go into some stockholders pocket or to fund unrelated good works.
As far as expenses go, we've detailed in our proposal a carrier-grade installation. That includes 24x7 customer support, and all the hardware/software bells and whistles that you expect from a central piece of infrastructure: redundant disk, switches, transit providers, UPS, and a set of installations spread throughout the key exchange points to insulate us from any one (or even many) localized or transit failures.
And, you may ask, why do we feel that it is necessary to provide this degree of stability? The short answer is because the criteria for evaluation of our bid say we need to provide this degree of stability. We're bidding to run .org because we think that .org is a public trust and that by running a truly open registry we can change the rules of the game.
Some think the entire process is flawed and ICANN should be abolished. Some argue that the entire registry function could be run on a far smaller budget. For example, instead of a full rack of equipment, one could slap a couple Linux/Intel boxes into a handy hosting facility and provide the same service.
We don't know if that's true. If our solution is stainless steel, then the Linux/Intel approach is plastic. In a few years, we think the $1299 registry-in-a-box commodity will be real, but for now the game is played with more expensive boxes. We know we can do the job with our people and that hardware and we know that a mini-boxen approach won't be taken seriously by the ICANN board.
So, it really comes down to revolution or evolution. We think we can make a big difference to the existing system. Any revolution is going to be a while in coming, and in the meantime we can make sure that at least part of the Internet infrastructure is being run in the public interest. And, to do this particular job means bidding carrier-grade equipment and support.
We should note that we're the only bid to publish our revenue and expense projections for the full period of the registry contract. We're not low-balling or gold-plating expenses or revenues: we're simply saying in great detail how we would operate the .org registry as an open and transparent public trust. Compare us to the other proposals and you be the judge.
| Comments |
|
 |
Subject: Re: Why Alpha
Posted by Carl Malamud at 07.10.02 @ 11.53 PM |
> Doesn't running a proprietary OS on proprietary hardware go aganst the grain of your proposal?
Hmmm .... I didn't think Unix was a proprietary OS. Tru64 is just a Unix variant. The reason we run that is the 64 bit hardware. When we run 32-bit hardware, we usually run NetBSD. So, your question is why we're running on the proprietary Alpha platform instead of the open Intel one? :)
We picked 64 bit for a variety of reasons, many of them having to do with the fact that this is the environment that we've used to provision other high-performance systems in the past. Intel doesn't have a viable and well proven 64-bit hardware architecture.
Why 64-bit instead of the so-called commodity approach? Both Google and Wayback (with lots of help from Amazon/Alexa) spent quite a bit of time tuning their systems to make it work on their rather demanding applications. With a small staff and a short transition schedule, we simply didn't have the big bucks that Brewster and Larry had to throw a pack of engineers into their systems to make them work.
Will we migrate towards 32-bit systems? Probably. Would that have been a smart bid to do a 3-month transition? I don't think so. Would a rack of 32-bit systems be more effective than a couple of 64-bit systems if you're trying to scale this up to 100 million names or do some really serious DNS response time? Again, we don't think so: our experience on the "F" root server was that it needed the throughput of a larger system.
In this bid to ICANN we had to promise stability above all else and do so in a short period of time with limited staff. This was the right solution. But, all our software is standard Unix on standard, freely available databases with standard protocols. That's the true mark of an open system. Why do you care if we run Tru64 if our software lets you run it on Debian, NetBSD, or any other variant? The important part is the open interfaces and the standard technologies.
|
Subject: Why HP Alphaservers?
Posted by Aaron Swartz at 07.10.02 @ 10.23 PM |
Why did you decide to go with HP Alphaservers running Tru64 instead of something like a rack of commodity Linux boxes? Both Google and the Wayback Machine have taken the commodity box approach and claim that it's the simplest, cheapest and most reliable approach. (How the Wayback Machine Works, Technology Behind Google).
Doesn't running a proprietary OS on proprietary hardware go aganst the grain of your proposal?
|
Building Public Parks for the Global Village Since 1993. 