5 ms·
I like this model. It's something I wanted to do a few years ago but of course didn't. I think this works best for complex and expensive computational APIs.
by TheDudeMan 5y ago
I like this model. It's something I wanted to do a few years ago but of course didn't. I think this works best for complex and expensive computational APIs. For instance, "do an FFT on these 2^N-1 data points" or "factor this 20-digit number". Things that are fairly easy to state in words, but tricky to implement correctly. Maybe you can't find a good implementation in your language.
Also, this is probably most useful for prototyping scenarios. In production, you want something faster and cheaper.
- shoo 5y agoFFTs or integer factorisation are pure-functional tasks and seem more suitable for a simple library call and local computation, without involving the network or adding new external dependencies -- assuming you have code running somewhere with sufficient compute. i find APIs that let you interact with or observe the world far more compelling (e.g. get the latex forex rates, make a phone call, etc), especially if they put a nice stable interface in front of a lot of complex or evolving details you'd rather not have to care about.
- dolmen 5y agoThe rule is: if you data is local, the code to handle it should be local. APIs that provide data are useful. APIs that just transform data that you provide much less because much less efficient and also bring integration nightmares (think not just today but long term).