Forced updates? Hmmmmm.

[Garlan Garner]

I know that I won't have to ask anyone's opinion about this.

I understand Tesla's position - and I understand an individuals position on this one.


Hmmmmmm.


https://electrek.co/2020/02/14/tesla-force-software-updates-owners-resisting-them/


Maybe I don't need an opinion. <------- that's an option.

 

[ewoodrick]

People who want the old interfaces can absolutely stay on them.

But progress is progress.

There are always those who can't move forward.

 

[dmurphy]

All I’ll say is ... I’m genuinely impressed that Tesla has a single stream of software that runs on any Model S, X, 3 or Y. Model S’s purchased in 2012 still get latest releases.

this is what a 2012 Mercedes looks like:

That lousy small screen interface with whatever it did then? It doesn’t do anything new. Just have to hope your Microsoft Zune doesn’t break.

Meanwhile, Tesla says “hey after 9 years, we need to make some big backend changes. Please update your car so we can keep sending you cool new stuff ... otherwise, you’ll end up like that Mercedes.

 

[Fresnel]

Maybe with these forced updates I won't have to tap the button to download and install the update.. That would be nice. A little more convenience for everyone.

 

[ewoodrick]

Maybe with these forced updates I won't have to tap the button to download and install the update.. That would be nice. A little more convenience for everyone.

No, that's not the case. This is NOT a forced update as you indicated. It's still an optional update.

What it is is that Tesla is changing some infrastructure. At one point in the near future, the old connectivity is going to be retired.
To end users, the GUI and the process won't look any different.

 

[Knightshade]

. At one point in the near future, the old connectivity is going to be retired.
To end users, the GUI and the process won't look any different.

well, for those who were sticking to especially old versions the GUI and process will look different

 

[camalaio]

All I’ll say is ... I’m genuinely impressed that Tesla has a single stream of software that runs on any Model S, X, 3 or Y. Model S’s purchased in 2012 still get latest releases.

this is what a 2012 Mercedes looks like:
View attachment 511101

That lousy small screen interface with whatever it did then? It doesn’t do anything new. Just have to hope your Microsoft Zune doesn’t break.

Meanwhile, Tesla says “hey after 9 years, we need to make some big backend changes. Please update your car so we can keep sending you cool new stuff ... otherwise, you’ll end up like that Mercedes.

Your point would be a good one if that's all Tesla was doing. However the statement is they're going to remove features if you don't update, not just let it stagnate in history like your comparison.

I'm fine with this as long as there's actual security-related urgency behind the updates which necessitates disabling those features. I've been there within a tech company. However if it's just an excuse to stop supporting old software (which I also understand and have experienced) I find that a bit less admirable, if still understandable.

Meh. I don't know enough to be mad or happy.

 

[ranova]

This probably affects users on V9 or earlier but Im still on 2019.36.2.4 by choice because of all the AP regressions introduced in the last few updates (ping ponging, taking turns too quickly, etc)

 

[boaterva]

It could be as simple as (I’ve seen mentioned) the certificates are expiring and you need to be on software x.y.z at a minimum to be able to connect to the mothership via the VPN. Who knows?

Tesla being Tesla they aren’t telling us the details.

 

[dmurphy]

Your point would be a good one if that's all Tesla was doing. However the statement is they're going to remove features if you don't update, not just let it stagnate in history like your comparison.

I'm fine with this as long as there's actual security-related urgency behind the updates which necessitates disabling those features. I've been there within a tech company. However if it's just an excuse to stop supporting old software (which I also understand and have experienced) I find that a bit less admirable, if still understandable.

Meh. I don't know enough to be mad or happy.

It’s not that they’re going to remove features, but stuff that requires something on the server side will break. It's unrealistic to expect there to be NO server changes for 9 years. That's not the way IT works. It's also totally unrealistic for them to go back and update every single version of prior vehicle software to entirely change the connection mechanism. You have to draw a line in the sand for such things.

Heck, I had a conference call with some very senior executives about exactly such a thing this afternoon on a very well known product. At some point, it's just not practical to support anymore.

I know what Tesla is moving from and to, and let's just say OpenVPN sucks from a scalability perspective. It's not practical when we start to scale to millions or tens of millions of cars on the road. So this move to Websockets is absolutely more secure, and necessary.
EDIT: Oh, I guess the Electrek articles doesn't mention *what* the backend changes are. I had read elsewhere that they're moving the cars from an OpenVPN connection, to one based on Websockets. That's a pure scalability-and-security play. OpenVPN can be a nightmare to manage, and even more to manage at scale. Websockets makes a heck of a lot more sense.

I get it - some people don't like change. But when you buy a Tesla, you have to think of it more as a computer than as a car... and in the computer world, things evolve quickly. To me, this is like saying "But I don't WANT to give up Netscape Navigator - so you can't change your website at all."

Time marches on, whether we like it or not.

 

[RayK]

To me, this is like saying "But I don't WANT to give up Netscape Navigator - so you can't change your website at all."

The way your conversation was going, I thought you were going to say "Mosaic".

 

[AdamVIP]

I'm okay with new stuff as I assume most here are but there's increased risk in forcing things. As time goes on things will change and there's more chances to brick something. If Tesla forces an update that bricks your car out of the warranty period it needs to be on Tesla to fix it. It will happen someday.

 

[Msjulie]

I wrote this in another similar thread: it's not inconceivable or even that difficult to support older API versions, companies do it all the time. Of course there is a cost in the quality testing bits and that's something Tesla probably doesn't want to do (some regressions suggest they could use improvements here)

I think there are other questions - if Tesla makes legally required changes to something like AP and you don't get the update, who is responsible?

I feel there must be other reasons beyond just the API versioning bit to why this is happening, perhaps it relates more to something has been cracked and they want to tighten up the client (car) side of things. Just guessing..

 

[dmurphy]

I wrote this in another similar thread: it's not inconceivable or even that difficult to support older API versions, companies do it all the time. Of course there is a cost in the quality testing bits and that's something Tesla probably doesn't want to do (some regressions suggest they could use improvements here)

I think there are other questions - if Tesla makes legally required changes to something like AP and you don't get the update, who is responsible?

I feel there must be other reasons beyond just the API versioning bit to why this is happening, perhaps it relates more to something has been cracked and they want to tighten up the client (car) side of things. Just guessing..

This change is WAY beyond an API. It's the plumbing altogether.

The previous configuration wrapped everything in a VPN.
The new configuration uses secure Websockets.

If it were just an API version, no big deal. But it isn't... this is everything about communication to the mothership that changes.

If you've ever had the ... pleasure ... at deploying OpenVPN at scale, this makes a whole lotta sense.

 

[Msjulie]

Where did you get all those details? Agreed it sounds somewhat painful

 

[Knightshade]

Where did you get all those details? Agreed it sounds somewhat painful

Green posted about it for one

green on Twitter

 

[ElectricIAC]

If Tesla hadn’t been chargegating folks with older packs in the interest of reducing warranty replacements, perhaps there would be less holdouts. Do the right thing and treat the early adopters with the respect they deserve because without their confidence in the brand at the time, there’d be no Model 3/Y and perhaps even no Tesla as we know it today.

 

[camalaio]

It’s not that they’re going to remove features, but stuff that requires something on the server side will break. It's unrealistic to expect there to be NO server changes for 9 years. That's not the way IT works. It's also totally unrealistic for them to go back and update every single version of prior vehicle software to entirely change the connection mechanism. You have to draw a line in the sand for such things.

Heck, I had a conference call with some very senior executives about exactly such a thing this afternoon on a very well known product. At some point, it's just not practical to support anymore.

I know what Tesla is moving from and to, and let's just say OpenVPN sucks from a scalability perspective. It's not practical when we start to scale to millions or tens of millions of cars on the road. So this move to Websockets is absolutely more secure, and necessary.
EDIT: Oh, I guess the Electrek articles doesn't mention *what* the backend changes are. I had read elsewhere that they're moving the cars from an OpenVPN connection, to one based on Websockets. That's a pure scalability-and-security play. OpenVPN can be a nightmare to manage, and even more to manage at scale. Websockets makes a heck of a lot more sense.

I get it - some people don't like change. But when you buy a Tesla, you have to think of it more as a computer than as a car... and in the computer world, things evolve quickly. To me, this is like saying "But I don't WANT to give up Netscape Navigator - so you can't change your website at all."

Time marches on, whether we like it or not.

Oh yeah totally, I do not disagree that things need to change over time. Like I said, I've been in this situation or similar before as well within a tech company. Sometimes it's a true urgency, sometimes it's more motivated by convenience and you need to screw over a few stubborn customers. It's the burden of the software industry as a whole, really.

I'm not super familiar with their OpenVPN to WebSockets transition (this is the first I've heard of it). On the surface that sounds like a massive reworking of their interfaces which probably came with slew of other changes either out of necessity or "well we're redoing it anyway!". If that is indeed the case, then they'd be even more anxious to stop supporting the old OpenVPN interface. On the other hand, maybe this is just making public the WebSocket interfaces that previously went over the OpenVPN link (if they already existed -- I have no clue).

It should be mentioned that it appears the security enhancement is not one directly for the customer in that case, but for their internal systems. Both are great, but we probably shouldn't assume (for others reading this thread) that this new system is more beneficial to your in-car web browser or phone-as-key security, which are completely unrelated.

This change is WAY beyond an API. It's the plumbing altogether.

The previous configuration wrapped everything in a VPN.
The new configuration uses secure Websockets.

If it were just an API version, no big deal. But it isn't... this is everything about communication to the mothership that changes.

If you've ever had the ... pleasure ... at deploying OpenVPN at scale, this makes a whole lotta sense.

Unless you know more than I do, there's nothing public to say that this new WebSockets approach isn't what was already inside the OpenVPN tunnel. If that's all that was going on, the project is a lot simpler in scope. OpenVPN implies there could have been more of course, but non-ideal solutions are chosen to problems every day in software.

 

[dmurphy]

non-ideal solutions are chosen to problems every day in software.

Technical debt is a very real, very painful thing. I don’t know where on that curve Tesla’s OpenVPN implementation is, but if I were a betting man, I’d say they’re carrying a heavy load for it. I don’t believe they’re just publicly exposing the Websocket interfaces here; I don’t believe the older firmware versions used Websockets at all. They used a different mechanism inside the tunnel. Remember, they implemented the OpenVPN solution back when they had less than 2,000 customers, total. Hindsight is 20/20; it was probably a good decision then, but never would’ve worked for Model 3. Having to keep two infrastructures and two sets of servers offering the same services carries a large penalty for a very small number of customers. It’s time to bring those forward.

anywho - having lived this life, I’d rather see them rip off the bandaid, stop carrying a huge technical debt load for legacy systems, and implement something modern and supportable.

I’m in the midst of watching my one and only customer deal with 30+ years of technical debt that’s coming home to roost. I guarantee you use their things ever day, and if you saw what it takes to make it all work, you’d never believe it. LOL!

So maybe I’m more sympathetic to what Tesla is doing here. It’s never easy to make these type of customer-impacting decisions, but the technology penalty for not pulling those forward is severe.

 

[Cyber_Dav]

I don't mind updates, but Microsoft updates (as an example I'm familiar with) have a bad history of breaking things. Yesterday's update broke my wife's Internet connection. She had to go to an earlier restore point to get Internet back.

Tesla has broken AP using updates based on an earlier AP version, so AP improvements were lost.

I just like to delay updates... usually until the "fix" to the previous update is released.