DJI Virtual Flight (iOS) has been broken for five months

I think they’re saying every hardware company is this way, with rare exceptions.

Sounds like the world in which the platform keeps moving and breaking APIs is a pretty bad world!

Yes, 100%. In places where they get software residuals (agriculture and enterprise, via FlightHub and other services), development is constant because a team stays assigned to it.

I do think that in recent years DJI have gotten way better at budgeting for after-sales support on mainline products. The Mavic 3 series have gotten consistent and meaningful updates for quite some time now. This Virtual Flight app was always kind of a joke and I’m not really surprised it’s abandoned.

> Shouldn't they already factor in the cost of app support for the next X years when selling the product?

Depends what you mean by “should”. If you mean that they should because it would be the right thing to do, sure. If you mean it would be competitively advantageous, well I rather doubt it. Consumers have almost uniformly demonstrated that price trumps all.

I suspect some of this is why more recent DJI drones have an OS and screen on the controller, because that way they have a single set of hardware to target, and even if some future Android release breaks backwards compatibility they can simply choose not to update to it and the controller will still work.

I think hardware companies just view programmers as, like, EE that aren’t even trustworthy enough to put in charge of keeping the magic smoke in the chips.

Thanks for the post about this. I have a Parrot Disco that I was planning to take for its first flight in about two years next weekend. Honestly these companies need to be sued over this behavior. Shutting down the servers without updating the app to work without needing login essentially bricks a product that I paid for.

Maybe a software-esque “right to repair”

The problem is that the software is most likely not composed by them in any meaningful way. If you take apart any proprietary firmware and controlling application you are almost certain to find boilerplate code provided by the chipset vendors (protected by NDA) glued together with other code from other vendors with a GPL violating Linux backend.

Recent example: I patched in to my 3D printer's mainboard using a TTL to USB and opened up putty and guess what shows on the console?

> Linux version 3.4.39+ (zhanglei@ubuntu Revision:543)

* https://pastebin.com/CPB7RBdK

That would be a great way to guarantee that nobody builds hardware in your country.

It would also do absolutely nothing in this situation where the product already comes from a foreign country.

I am just imagining this, or did it feel like 10-15 years ago breaking changes were a big deal, and companies went out of their way to provide an upgrade path. Now my flutter app from 3 months ago won't even compile.

Exec orders aren’t the same as culture issues among workers

The that cultural difference causes some pretty important secondary differences too - for example, Microsoft's way enables software to be a product - written once, working approximately forever; Apple's way pretty much forces a subscription model.

The Microsoft way is absolutely better than the Apple way. Platform holders, Apple in particular, reap enormous profits. They have at least some basic responsibility here. It's hard? Tough shit, software is hard. Start putting in the work for your 400k TC.

This is also why Apple will never be a long-term gaming platform for anything other than gacha pull and otherwise exploitative games. I can still play the original Dark Souls with DSFix on modern Windows (or Linux via Proton for that matter). The game is timeless and should never be lost to lazy platform holders.

Yeah, this was well explained by Joel Spolsky in this essay from 2004

https://www.joelonsoftware.com/2004/06/13/how-microsoft-lost...

starting at § "The Two Forces at Microsoft" where he describes the Raymond Chen Camp vs. the MSDN Magazine Camp.

Audio software tends to experience breakage even after minor iOS releases. For example, Apple have once changed the in-memory format of MIDI sysex messages passed to the app in a dot release. Segfaults ensued. The particular callback wasn't documented well and some OS behaviors had to be inferred. Apple still went through with the change, not noticing or maybe not caring about consequences.

The PX4 stack would be where I would start if I were in the market to buy today. They have a list of compatible hardware and drones, including some relatively off-the-shelf drones. The list of drones provides a lot of detail about the various levels of support and functionality available.

There is nothing that comes close to DJI for a flying camera, and I don't say this as a fanboy but rather as someone who has been in this industry in various capacities for a long time.

Ardupilot and PX4 are great projects, but they're just a basic ingredient in making a drone with good autonomy. They handle the basics of stabilizing a multicopter or plane (with a lot of tuning required) and have some rudimentary range finding and optical flow support, but don't offer much in the way of state-of-the-art vision based autonomy for position hold, obstacle avoidance, or route planning. Even in the commercial space (where a lot of stuff is based on PX4, it's BSD licensed so products don't need to be source-available), nothing really gets close to DJI. On the other end of the hobby drone space, INAV and Betaflight are fun, but they're mostly for FPV which is a completely different ball game.

The gaps for full open source _and_ non-DJI commercial drones are numerous. Really they're pretty much end-to-end:

* Vision autonomy. There are a _ton_ of UAV vision autonomy projects on GitHub, mostly because it's a popular thesis project. Unfortunately, none of them have taken off as a practical, well-maintained thing you can integrate into normal drone hardware and use. In the commercial space, Skydio finally got a good win here, but they've given up on the consumer market. This isn't just a critical feature for whiz-bang inside-out pathing like Skydio do, it's an important fundamental for strong position hold (called visual odometry) as GPS is drift-prone and traditional low resolution optical flow struggles once a drone is at a higher altitude.

* Camera image sensor quality. This is the place where Skydio really fell behind DJI in the consumer market. This is a highly proprietary, closely guarded space. It's almost impossible to find a good image sensor plus the parameters to make it look good (ISP firmware / sensor parameterization) that's not locked under 100 levels of NDA. Add on the need for specialized optics and you're in for a really tall order. Even the Pi Foundation essentially gave up on this and ship a DRMed camera with proprietary ISP parameterization.

* Video downlink. This is a big one too, and one of the other big places where Skydio fell behind. DJI and Autel have access to modified mobile phone IP via China. Everyone else is using either analog video or WiFi and it just isn't good.

* Camera gimbal quality. Ardupilot did a GSoC project around this recently. This space is really underinvested too. Making a good gimbal is _hard_, and it goes hand-in-hand with the camera sensor issue in some ways because making a gimbal which is weight/balance-matched to a specific camera is a lot easier than a general purpose one. Also, most open source projects spent a long time focused on on bottom-facing survey / photogrammetry use cases because they were easier, so basic features like "aim/zoom to a tapped point on a video" aren't fully baked at all.

* Battery quality and parameters. Open source / hobby drones tend to use either polymer-electrolyte prismatic pouch packs with extremely high instantaneous discharge but comparatively low overall capacity (FPV drones) or round-cell lithium ion batteries with good overall capacity, but poor instantaneous discharge and poor packaging. The optimal battery for a small camera drone is right between these two extremes.

Even in the commercial space, I'd recommend DJI. Autel is basically one generation behind DJI at a higher price point. Skydio drones were cool but they exited the consumer market, and they were loud and had bad cameras and a weak video link compared to their contemporary DJI rivals. Parrot are interesting as a non-Chinese option, but they're about two generations behind DJI, also at an even higher price point.

If you want a camera drone today, buy a DJI drone with a standalone remote controller (the one with a screen). That way, you don't need to install their sketchy app on your phone, you can firewall the controller however you'd like, and you only need to give up limited information as part of the initial activation/login process and then you're good to go.

This behavior isn’t unique to Chinese companies

> to collect your personal and flight data.

Its no secret they save flight data (like everything you input, temperature, gps points...) in their cloud.

I think there are workarounds for this, a friend of mine never paired his drone with a smartphone connected to any network. Iam pretty sure this will not work for a long time.

Its very daunting if you ask me, the amount of data they gather. Not a small part of drones used in active wars, like ukraine right now, is by DJI. This is alot of power in the hands of a few.

It's literally DJI's own stock remote that comes with Mini 3 Pro, I haven't changed anything.

I own a HEXA robot, I was a kick starter and paid a price for it.

When I first purchased the robot it never required phone GPS. And that now if you disable it at any point it disconnects you from the robot. It sits gaining dust because I wish not to use an IOS app with GPS enabled for no reason.

Questioning support to why this is required they said "people are happy with GPS on" and that was that.

I look at it sadly because it runs Linux, has full root access and so much potential. But, I just don't like the fact that could be potentially tracking GPS coords back to HQ.

Nope. They have apps on the iOS App Store just fine. DJI Fly for iOS was last updated 4 days ago.

DJI used to have the "Mimo" app in the play store. It was rated very low, 2.3 stars. The app required users to allow the permission to read the phone's IMEI. If you denied that permission, the app closed. Myself, and I'm sure plenty of others, have reported this to the play store as something against the play store policies. Perhaps Google gave in to the complaints eventually and told DJI, enough is enough with your crap, who knows... And that made DJI say, "enough is enough with _you_, we're big enough to distribute these outside the play store and keep our invasive tracking and permissions".

Page visible only if you have installed the app in the past: https://play.google.com/store/apps/details?id=dji.mimo

This is the best advert for not buying an IOS device I've heard in a while....

never had any such problem on android, I still have working apps that target android 6...