Spotify's numerous "radio" channels

The service was able to set timers and alarms, and answer questions about summary weather forecasts, sunset times, word definitions (Alexa used the Collins English Dictionary for this), city and country distance / population / time, calculator answers etc. It also provided a facility to place voice calls to other Amazon devices once these were added as contacts, although the software sometimes gets itself into a failure mode where the "call" command is acknowledged but does nothing---power-cycling your devices may be needed when this happens. It used to provide email access too, but this was removed, so I doubted that the service to place voice calls to non-Amazon phone numbers in some countries---which obviously costs Amazon---would be provided for long.

Third-party functionality was available by selecting from what Amazon called "skills" which have to be "enabled" on your account. For example you could enable BBC Sounds and use it to listen to live or catch-up radio, if you remember to start your request with the words "ask BBC Sounds to play" so it doesn't go to the default track-request skill instead, which is either Amazon Music or Spotify. (There is also a failure mode where BBC Sounds gets stuck playing at a very low bitrate until your device is powered off and on again.) Some 'skills' were not available unless you opened your Amazon account in a particular language, for example Radio Swiss Classic could not be enabled unless you had your account set to German (even though you weren't expected to interact with it once it was playing), and some small developers even set their English-language 'skills' to be available only in America and not Britain (presumably by writing en-US instead of en somewhere). Developing a 'skill' essentially involves implementing and registering a server-side API (which Amazon steers you towards paying to host on AWS) but there'd be copyright implications if you provided music to them.

When I first got the device, it was possible to ask Spotify to play a particular piece of music and it would play it. But this functionality did not last. Soon, Spotify was interpreting all requests for a piece of music as a request for a virtual radio station I'd never heard of, with a name very similar to the piece of music I'd asked for but which was usually playing something else in the same general genre. At first I thought this was some third-party bad actor trying the equivalent of 'typo squatting' by registering their "radio stations" under the names of pieces of music, so you couldn't ask for the piece without getting the station, and I tried to work around it by asking for obscure early-music recordings but it seemed the "radio station" people had hijacked even those. However, when I carefully read the description of the Spotify "skill" and translated their marketing language back into normal English, it became clear that, whereas previously you could ask for a specific piece of music without charge but might occasionally be required to listen to an advertisement, it was now the case that you had to pay extra for the privilege of being able to ask for a specific piece of music, and if your account was not paid up then Spotify would generate a "radio station" to play something in (what Spotify thinks is) the same genre but not exactly what you'd asked for. It's possible that Spotify gave us a trial run of the exact request service and the trial expired, or perhaps Spotify's business model changed; it was not clearly explained. (At least the Amazon Music service did explain that specific requests now required payment.)

Spotify's "radio" algorithm seemed to be going out of its way to be indirect: even when I asked it for anything conducted by Christopher Hogwood (thinking I'd at least get period performances), it played three random tracks from non-period performances of Bach, Purcell and Handel, then some 'mood' music (I think it confused Scott Ross the harpsichordist with Scott Ross the modern composer), then a bit of Biber, then three advertisements (one of which was for gambling) and a repeat of the same Bach track before finally playing part of an overture that really was conducted by Hogwood, after segueing into a movement of Telemann and then back to the Bach (it managed to pick a different track this time) then a modern improvised prelude from a Carl Friedrich Abel CD and a chorus from a non-period performance of Purcell's King Arthur. If you want any more control than that then you'll need an extra subscription. This of course makes the basic service completely unsuitable for those "don't you know that piece? let me play you that piece" conversations.

On the other hand, the specific-request service that we'd previously been given didn't always work either---it was liable to do things like play the Mozart Requiem if asked for a specific Mozart concerto. (Sometimes I miss the 20^th-century search systems that could say "no, we haven't got that" instead of pretending you'd asked for something else.)

Using Bluetooth instead

There can be setup issues---when I tried to 'pair' it with a Raspberry Pi Zero W, the pairing failed, and a mid-2011 Mac Mini paired but couldn't keep a connection open reliably enough to play audio to an Echo Dot 3 but did play to an Echo Dot 2 (but it was not able to play to an "aggregate device" of two of them even though it could play to either one, and a little 'choppiness' was sometimes in the audio, and if it had been connected too long it might fail to play until rebooted). I don't know enough about the lower levels of the Bluetooth protocol to diagnose these differences in behaviour.

A 2019 MacBook Pro was able to use two nearby Echo Dots as a stereo set---in Audio MIDI Setup create an "Aggregate device" with left assigned to channel 1 and right to channel 4, turn down the 'wrong' channel on each device and increase the overall volume. This did not work to another room though (range issues).

Pairing to a single Echo Dot worked from Amazon's own Kindle Fire tablet and from a Samsung S21 phone, which can also pair to two devices at a time, but cannot turn down the 'wrong' channels on each device (and neither can the Dot itself---at least not unless you connect an external speaker and rig it to get only one of the two channels), and pairing mostly worked from a Samsung S9 phone and a Raspberry Pi 400.

If you do make Bluetooth work, you can still ask the Echo to change the volume or to answer simple questions (like ``what is the time'') while Bluetooth is connected, but if you ask it to play other audio then it will disconnect from Bluetooth first and you must reconnect if you want it back later (it seems there are some exceptions to this depending on which "skill" is playing the audio).

You will not be able to combine Bluetooth input with the Echo Dot 3's "stereo pairing" function instead of driving each device separately: the "stereo pairing" is reserved for the Dot 3's own audio.

You can temporarily suppress the Bluetooth "now connected to" messages by turning off "Announcements" in the compulsory Alexa 'app' under the device's Settings / Communication, and rebooting the Dot. But this does also have the effect of exempting it from manually-initiated site-wide announcements, and the "connected" messages can reappear after a few hours, plus the beeps are still present.

There are also third-party 'skills' which can supposedly add the facility of playing publicly-accessible audio from arbitrary URLs (which Sonos can do in its default API but not Echo); however I was not able to make any of these 'skills' work in practice---my attempts to place RSS, Atom or MP3 files on an external server and add the URL to a skill did result in some AWS machine fetching that URL, but the Echo kept saying "I can't play that right now" with no technical information to give me any clue about what changes would make it acceptable.

The above is based on my experience and is not legal advice.