https://www.ted.com/talks/clifford_stoll_the_call_to_learn?l...
I'd prepared a 1 hour talk. When I was about to go on stage, they told me that I had only 15 minutes. So, well, I remembered one of the few things I learned in grad school: talk fast and don't give 'em a standing target.
Covered my points and finished in, yep, 18 minutes.
I still have the hat. I no longer have the bottle as I moved to the other side of the world a number of years ago, and after inquiring with some mathematician friends it turned out that four-dimensional glass bottles are not compatible with three-dimensional backpacks :-/
When I got the package it was covered in hand-written notes; gosh, I don't remember what it said exactly, but it was hilarious; somehow you even managed to put some joke in Dutch on it. I wanted to email something back but I was shy and didn't know what to say. So, about a decade too late, thanks! Not only was I very happy with the bottle, the packaging absolutely made my day all those years ago.
Thank you for what you do.
I still don't understand how it works. Why can somebody owning something called "Amvoom" claim something called "Acme Klein Bottle"?
An even if they legally own the brand, how keeping the reviews when moving the brand to the new owner is the proper thing to do for the customers? By definition, the reviews are for another provider. I don't get it.
I really hope I'm wrong though because this sounds like a very lazy and flawed system.
That's how you make $100B, not by "doing things that don't scale" like handwriting thankyou notes to every customer.
Sounds like most of his predictions in it (eg e-commerce will fail, digital books will not be viable, etc) were wildly off the mark - but were any prescient?
For example, he complains that nobody will want to look up information from computerized databases because CD-ROMs are too slow; consumers won't shop online because they can't pay securely; retailers won't rely on e-commerce because too few customers have Internet access; nobody will want e-books because you can't read them on the subway; there's no way to effectively search for content online; digital art will never surpass clip art and crude photoshops; it's impossible for networks to be secure because data and credentials are unencrypted; and so on.
On the bright side, he thinks that at least nobody will need to worry about online privacy, because it will be too cumbersome for anyone to effectively maintain databases of personal information.
But on the other hand, with some of his observations, it's at least arguable that they still hold true 25 years later:
> Anyone can post messages to the net. Practically everyone does. The resulting cacophony drowns out serious discussion. Online debates of tough issues are often polarized by messages taking extreme positions.
> An original IBM PC, now over ten years old, is fully obsolete. Likely, it will still work perfectly and do everything it was build for; after all, the silicon and copper haven't deteriorated. But you can't get software for it any longer.
> A word processor may last two years before the next version. These upgrades likely add as many new bugs as are patched, and result in a bigger, more complex program. One that's less and less compatible with old files. [...] Curiously, as computer hardware gets faster, programs run slower.
> Photo retouching isn't new. Digital image processing, however, can be so extensive yet undetectable that it undermines the foundation of photojournalism -- that seeing is believing.
But he was right about tech will enable and amplify the worst in social behaviors.
Here's a video of him describing how the Curta performs arithmetic: https://www.youtube.com/watch?v=OynMJB-2J1o — as well as the SciAm article he wrote on them: http://www.mycurta.com/Calculator.pdf
Do you have a caching service in front maybe?
Yep. Cloudflare is out front, so the actual load on the rasp-pi is mitigated by their content-delivery network.
Then, too, my website is almost entirely simple html with compressed images, so there's not a lot of bytes to shovel.
Here in Berkeley/Oakland, Sonic.net has strung quality fiber-optic, so there's 1Gbit to my house. That lets me keep up with things. However, they only give a dynamic ip address;, so my pi must keep track of its address and tell Cloudflare whenever it changes.
Works surprisingly well - from /top/ I see about several dozen simultaneous users (thank you!), and the cpu temp is about 2 degrees above its normal of 50C
The raspberry pi itself is in the crawlspace under my home, fed through a Ubiquity edge router. Much fun, playing with Unix (oops, I mean Linux) -- sends me back to days of yore when everything happened from your command lines.
1. I met you in Kepler's when Silicon Snake Oil came out and we talked about something and you wrote in the inside "I hear you, John". I don't know what we discussed!
2. I am now Cloudflare's CTO and if you want to avoid the dynamic IP address problem you can use Cloudflare Tunnel to connect to us (rather than us to you). https://www.cloudflare.com/products/tunnel/
And that Kepler's talk? Happy memories, indeed. They "paid" me for my talk by saying that I could have a copy of any book in the store. I chose the Times World Atlas (a way-big book of maps). The manager's face suddenly dropped -- and then I told 'em that I'd pay full list price if all of their employees sound sign the book. Result: I now have a terrific atlas of maps, with a dozen signatures of book people. (two of them visited me last year and I showed them their signatures from decades ago -- very sweet!)
Meanwhile, I gotta send out some of the tsunami of Klein bottle orders. But Cloudflare tunnel? Here I come!
My needs are very similar to Cliff's. I have (or going to have) a home server on a residential ISP which can't accept incomming connections. I need a public-facing server that will return from the cache if I'm ever on the HN front page or my home server is down, TLS handling, and routing to my home server's port 80 when needed.
I mostly serve personal websites but I have aspirations of building a small business. Is there anything in Cloudflare's range that would work for my kind of needs?
Hoping you follow-up on nine-day-old threads, Bill.
2. We literally announced free tunnels for all customer here: https://blog.cloudflare.com/tunnel-for-everyone/
The blogs that go down here typically back every request by MySQL (ahem, WordPress) which is totally unnecessary and often actively harmful since MySQL has very low default total connections allowed.
The point being: don't serve requests backed by a database unless the results are likely to change very dynamically!
WordPress is not my favorite thing and some of the available plug-ins do terrible things with MySQL, but the problem is not too low default connections; it's too many PHP workers. WordPress is generally focused enough that most of the wall time is spent in waiting for the database, so you want to optimize for throughput; one or two workers per cpu thread is plenty for that. More concurrency than execution available reduces throughput, so it's better to queue requests in your http layer than to process multiple at once.
Large numbers of MySQL connections are more appropriate when the web pages do a mix of things, but more/mostly idle DB wise; in that case, you might still want persistent connections to reduce round trips before a query, but are less likely to have a query backlog large enough where task switching overhead becomes significant.
The only difference between this and adding a cache is that the cache is another piece of software in your production stack.
Vanilla SSGs are so simple I ended up writing a basic one out of a markdown and jinja parser in Python every time (for example: https://github.com/eatonphil/notes.eatonphil.com/blob/master...).
If I were not lazy I might learn one of the major ones like Hugo.
It doesn't matter what SSG you pick, they all produce the exact same kind of thing.
But the point was to make a comparison to a Raspberry Pi and emphasize that you do not special compute to withstand thousands of page views. Even S3 and GH Pages are overkill in terms of the compute behind both of them vs. what you need minimally.
And I will double up and say it is a great episode of a great podcast.
Out of interest, is it publishing companies that make digital versions of books, or is it up to the author themselves to do that?
(And if it's the latter this is a small request to Clifford to consider it if it's not too much of an arduous task. And provided their relationship with amazon isn't too soured by the current situation, which I could understand if it is).
Also, if anyone is unaware, this is this Clifford Stoll: https://en.wikipedia.org/wiki/Clifford_Stoll - who wrote this brilliant (and true) book: https://en.wikipedia.org/wiki/The_Cuckoo%27s_Egg_(book) - which is a really good read and perfect for HN.