It used to be that you could buy a computer, run software on it an it worked, and it stayed working forever.
I saw a guy at a hertz rental place in the early 2000’s running the truck rental service on an ibm pc xt with a green screen and an epson dot matric printer and everything and it all just worked.
But now everything is on the internet.
So you can no longer be sure that any software you have will continue to work as long as your computer does because it has to talk to other computers that might be changing in some way.
This happened to me (again) today.
I have a machine that has a dynamic ip address and it uses inadyn to update the ddns server at afraid.org so that I can find my machine when I want to.
But the dynamic ip changed at some point and I could no longer get to the machine. I eventually got access to it and found that inadyn was no longer able to talk to the ddns server:
“The response of DYNDNS svr was an error! Aborting.”
Like all good software it told me what the problem was in detail and how to fix it.
After some rummaging around I found that the server inadyn was talking to was no longer supporting the old version of inadyn I was using, there was a new protocol and my old old version didn’t speak it.
Because I’m attached to the global supply chain, I can no longer expect things to remain working, because some other parts of the global supply chain might change and I’ll have to change as well to keep things working.
So I said, okay, I will fix it.
I downloaded the latest version of inadyn and tried to build it.
inadyns requires libconfuse.
So I download that, and try and build it, but it requires gettext.
So I build that. That works.
Then I go to build libconfuse again and it fails with some lexer problem.
I download a version that’s a bit older than the newest and built that, it builds.
Then I go back to inadyn and it builds too.
I install it, and run it, and it says… it requires a newer version of gnutls than I have.
So I download gnutls and try to build it.
It says in requires nettle 3.4. So I download that and build it. It builds.
I try to build gnutls again and it says it still requires nettle 3.4.
I google and there’s a few answers on stack overflow, but none solve my problem.
At this point I stop and I wonder what the purpose of all this is.
Somewhere at the bottom of this chain of rabbit holes I expect there will be a circular dependency making it impossible to get working.
At this point some of you are wondering “what kind of machine is this that you can’t just use the package manager to get the latest package.” It doesn’t matter, that’s not the point. It’s all broken. It’s a pile of houses of cards stacked on top of each other.
One of the more amusing points was when I noticed libconfuse titles itself thusly: “Small configuration file parser library for C.”
I am libconfused as to why it takes so long to build something small and why there are so many source files just to parse a config file. Or maybe what they’re saying is that it can only use small configuration files, and large configuration files are beyond it. It still shouldn’t take that long to build. I have a small c++ class I use to read config files. It’s about 100 lines long. I can compile it faster than I can hit the period at the end of this sentence.
If it’s 2019 and we can’t make this a simpler process, then maybe it’s not worth doing at all.
But it doesn’t matter whether I like it or not or whether it works or not, because we are all part of this global interoperable supply chain that now requires you keep up to date or no promises that it will continue to work.
For really important things (ie, systems where money is traded) apparently there’s some notification system to alert you to impending breaking changes, but for anything that isn’t about transferring money, you just better keep up all the time, or suffer unexpected compatibility failures when somebody else decides to break something you set up years ago and left running because it worked.
I agree about things breaking in the global ecosystem and it sucks, but as for your dynamic DNS problem, you don’t need a client at all. afraid.org accepts updates using HTTP. Here’s what I’ve got in my cron.daily:
/usr/bin/curl http://username:password@freedns.afraid.org/nic/update?hostname=yourhostname.example.com
…which makes one wonder why the series of rabbit holes that is compiling inadyn even exists in the first place.