Yes, the exact same snarky comment that comes up every time someone brings up the word "serverless." Anyone who is using this code base is probably smart enough to connect the dots that yes, the code is running on some server some where.
When it was first came to my attention, I assumed it meant "peer to peer" or "all code runs locally". I don't feel like that's because I wasn't smart enough. It is a fairly silly name.
I don't understand the hate. You know what it means and are just being a jackass. Anyone who knows what it is knows there is a server somewhere being managed. It has a valid purpose and for all intents and purposes you as a user have no interaction with server level details and are delegating all of those details to some service provider. So from a user point of view, it is "serverless".
In a profession which depends on knowing the meanings of words and symbols and using them exactly correctly, naming a software deployment model as a thing that it is not is fundamentally wrong.
Why is it wrong? Because everything depends on you having a model in your head of how the universe works. When you persistently misname something in a way contrary to reality, it becomes part of your model. "It's serverless, therefore I don't have to think about servers in any way."
From a user's point of view, everything is magic. That doesn't justify calling software magic.
Does Docker or Centos or Beanstalk or Kibana mean anything? No, they're just symbols. It's just a name and you have to know what it means. Everything in tech has some stupid name.
You're not missing any part of using serverless by not caring about servers. You have platform constraints to minimize.
How? Pointing your resolver to something returning whatever has been possible since DNS was invented. This changes nothing except that your ISP isn't able to read/change responses.
privacy, which is essential for security. currently your DNS requests are sent in the clear. and even if DNSSEC sort of guarantees the integrity and authenticity of the answers, it doesn't provide secrecy.
and yes, currently TLS (+SNI) leaks the domain name (server name in TLS parlance), but there's work being done on that and it's very likely that in a few years that leaky hole will be closed too.
That depends on how close you're looking at the tail. Things like ISP servers that purposefully or not corrupt information, caches that cache things for weeks after the ttl expired, ISP servers that simply don't respond in a timely fashion. It's pretty useful to have an alternative way to resolve hosts.
24 comments
[ 2.9 ms ] story [ 70.3 ms ] threadIt's the current "web scale".
Why is it wrong? Because everything depends on you having a model in your head of how the universe works. When you persistently misname something in a way contrary to reality, it becomes part of your model. "It's serverless, therefore I don't have to think about servers in any way."
From a user's point of view, everything is magic. That doesn't justify calling software magic.
You're not missing any part of using serverless by not caring about servers. You have platform constraints to minimize.
I have no idea what software or hardware my DNS provider uses. All the scaling, fault tolerance, backups etc are out of sight and out of mind.
and yes, currently TLS (+SNI) leaks the domain name (server name in TLS parlance), but there's work being done on that and it's very likely that in a few years that leaky hole will be closed too.