I would be very interested for as many people as possible to test this and give feedback.
Particularly testing with different postures, hands, walking, different sensors etc.
Very simple to setup and does not require expensive hardware.
AWS is not value for money, I do have a DO account that is great but my development is mostly hosted locally with tunnels from cloudflare, it is remarkable how far you (I) can get with that setup.
Last I saw, AWS has way better peering agreements than DO. Lots of problems with terrible throughput and lots of dropped packets for various clients (in several cities in North America, not just overseas or in the middle of nowhere) that vanished instantly on switching to AWS (including overseas ones that were also having problems)
Unfortunately, this isn’t something that shows up on spec sheets when you’re choosing a service. :-/
Source was we used it, and that's what we saw, ~20% of clients on three continents (about half in North America) consistently had terrible connectivity to DO (not none, but it was really bad) and we spent a lot of time trying to fix it. Vanished through nothing but shifting that to AWS. It was clearly DO's peering network.
You probably won't see this unless both the following are true for your situation:
1) You have a workload that makes this issue noticeable. Long-lived connections and large transfer sizes make it more likely you'll notice. Loading 20kb of static html over the connection likely won't seem to have any problems (unless you run repeated trials and network analysis tools). Of course, modern websites can be pretty large...
2) Your users are long-term enough and in communication with you so these issues can even be noticed in the first place. Also helps if they're technical. If you're not hearing the story and aware of the situation on the other end of the line, all you see is a slow connection, could be anything causing it, and there are plenty of them for reasons that have to do with things closer to the client's end.
So all e.g. an e-commerce site might see is a somewhat higher bounce rate than necessary (due to some fraction of their users experiencing the site like it's on a somewhat-jittery ISDN line) without even knowing they're leaving money on the table because they likely have no way of even being aware of the problem.
[EDIT] Yes, we tried shifting around a bunch of ways on DO's side trying all kinds of ways to fix this, I'm quite sure it wasn't that we were unlucky with our hardware draw there or just one of their datacenters had this problem. It was something past the edge of their network.
I suspect that you used highly optimized algorithms written for python, like the vector algorithms in numpy?
You will struggle to write better code, at least I would.
Python 1.4 would be mid-late 90s long before numpy and vector algorithms would have been available.
I suspect it’s more likely to be something like passing std::string by value not realising that would copy the string every time, especially with the statement that the mistake would be hard to express in Python.
We live and learn.
Still a huge fan though.
reply