Yeah I always recommend beginners start with an analog scope. It's a bit like learning how to fly the plane before you start using the fancy autopilot system. An analog scope forces you to understand what you're actually doing and then later the DSO makes many of those tasks faster, easier and more accurate.
Lets face it, analogue oscilloscopes are a 20th century artifact (*). Some old timers have a difficult time with the transition and lament the absence of this and that feature their high end CRO offered decades ago and now missing on entry level digital oscilloscopes. Not sure, whether there is real benefit in burden newcomers with that.
Quite some effort has been put into making electron beams reflect a given signal, with some amazing results, but lets not forget that it's just a means, not an end itself (except for collectors of antiques).
Whenever extra-sensory signals are transformed to allow humans to perceive them, some compromises have to be made. CROs make different compromises (and compromise they do!) than digital oscilloscopes. I think the latter are held back by the market expectations formed by the former. It's perhaps comparable to film vs. digital cameras. For a long time professional digital cameras used a mirror like the analog forefathers (often even the same body, just with a sensor placed where the film used to be). Cheap consumer cameras dumped the superfluous (**) mirror long ago, which then absurdly became a differentiator welcomed by the marketing folks. Only lately one finds high-end photo cameras without mirror.
Likewise I'd like digital oscilloscopes freed from the clutches (and crutches) of the Xray generating past (***).
And yes, in the beginning, the quality of the output of digital cameras was no-where near the one of film cameras, still they won, chiefly because they were good enough for the dominating use case (on-line photo sharing rather than poster-size prints) and their outstanding convenience and instant gratification.
The killer feature for oscilloscopes is the 'single shot' data acquisition. Yes, for reasonable quick signals it is possible in an inconvenient, time consuming and very limited sense with CROs (+ camera), but only digital storage made it accessible to most.
Digital oscilloscopes wooing consumers used to operate CROs mimic their user interface. But oscilloscopes aren't cars, a break in the user interface isn't a safety risk. New users are more likely to tinker with Arduinos and similar MCUs, than repairing TVs. They won't be averse of having a computer on their bench, in fact, their desk might be their bench. USB based (fast) data acquisition systems with protocol analyzer might be better suited to their needs than a 'bench' oscilloscope with knobs. Then the detour via CROs becomes even more questionable.
just my 2c
(I do use a CRO on my bench, chiefly because it has no fan, but for serious trouble shooting I use my DS1054z, for 'exploratory' tinkering, I actually prefer my Analogue Discovery, as it allows me to use a big screen, its 14bit ADC - so I can use it as signal and network analyzer - and the lack of a fan)
*) I am aware that there are some moderately fast CROs cheaply available on the second-hand market (e.g. TEK475) , which only quite expensive digital oscilloscopes can match, but I don't think there has been any new CROs developed in this millenia other than low bandwidth (<25MHz) models at rock-bottom price for the educational market and low income markets
**) Early electronic displays didn't offer the contrast or low latency professional users demanded, hence the optical viewfinder (necessitating a mirror or prism) remained (remains?) popular with that group for a long time. Further early sensors were self-heating during continuous use, which increased noise. Recent sensors allow fast, lower resolution modes with limited self-heating.
***) It's difficult to change the deflection of an electron beam greatly very rapidly, hence the CRTs used in CRO tend to be quite narrow and comparatively long. There was no easy way around it, so small screens became accepted by the market, but why should digital oscilloscopes be limited to a 7" screen? Even professional CROs offered often an uncertainty no better than 3%, which then resulted in digital oscilloscopes getting away with 8bit ADCs (whether higher accuracy is a valuable feature in an oscilloscope might not be so clear, but a higher resolution would allow for a more sensible FFT feature).