|
Thanks! It's great to read the backstory of how the original Spotify client came to be: > "Not to go into too much detail here, but at the time, most of the internet was made up of “thin clients,” like web pages or Flash-based clients that ran in-browser, and used more traditional, standardized protocols like HTTPS. Seeing the limitations of that, Ludde and a team of engineers ran in the exact opposite direction, creating a stand-alone “fat client,” building entirely new protocols and hybridizing client-server and P2P technology to suit their own ends. (Check out Episode 01, “How do you steal from a pirate?”, to hear more of that nitty-gritty stuff about persistent TCP connections and how our P2P implementation saved us bandwidth cost.) It was only by rethinking every layer of our infrastructure that we were able to pull Spotify off, to create that magic moment of double-clicking on a new song and having it instantly play. And speaking of magic …" Having been an early user of the beta I presumed it had to be driven by people with this mindset from rethinking everything from bottom up to provide the best UX possible, would really love to read more about the technology used in the original Desktop client? EDIT: Currently listening to “How do you steal from a pirate?” which is providing a more detailed backstory on the origins of Spotify: https://open.spotify.com/episode/1jHRUXkeiUh44CK4KZQb0h?si=d... Sounds like Ludde Strigeus, the creator of µTorrent was the key hire to make the original Desktop Client UX possible whose Desktop & P2P expertise was able to convince the rest of dev team to go down the path they did. Some interesting insights, they used Ogg Vorbis instead of mp3 using a custom designed TCP protocol because they were better able to strip packet bits down to transport just the audio bits required for playback. > Michelle: The thing that happened that was kind of pure magic in that meeting was that [Daniel] did a comparison. He started playing a song on the software, and the song played so quick, so instant … I mean, I don’t know if people remember, but playback was slow back then. Even if you had an MP3 on your computer, and you played it via, you know, Winamp, iTunes, this was faster. And we were like, “You have the files on your computer, right?” And he was like, “No, it’s in the cloud.” I had the same initial experience where I was blown away at how instant and responsive it was, at first I didn't believe it and thought it was doing some sort of pre-caching magic where it'd start downloading before selecting each song. So ran lots of tests where I did first time searches and immediately scroll to songs down the list of search results (to bypass any caches) and could see that it was indeed pulling traffic in real-time, the time from click/scrubbing to audio playing was just unbelievably fast. |
This Electron crap is really a race to the bottom, so you can hire the cheapest college drop-outs to cobble together some JavaScript to add feature number 1001.
On another note, this ludde guy is one of the few rock stars in IT to me. Had a lot of fun on ScummVM, played OpenTTD to death, used µTorrent in College, you name it...