Yep, many of the PCIe 4.0 NVMe drives that I have seen have a built-in heatsink - and you would have a hard time fitting that in the Librem 5. (So you would need to select the ‘no heatsink’ variant and the phone designer would have to find a way to dissipate all that heat.)
Still, in some climates, the heat dissipation could be seen as a ‘feature’. 
I wonder if it would be possible to have a smartphone with an easily exchangeable screen and to change between e-Paper and a normal screen. Just a crazy brainstorming.
sounds fancy. 
I just still think that developing a perfect completely multi-purpose notebook in smartphone factor is waaaaaaaaay toooooooo ambitious task.
So if there are users that want a more limited purpose device most of the time - a secure communicator device with e-mails, encrypted messaging, encrypted calls, and an SSH terminal, mp3 player ect. the e-paper screen would be a great way to consume less energy and keep the Librem 5 alive over longer periods of time.
And if the user says that tasks that require high refresh rate and colours like video calls, watching videos and playing games are essential, then these could use the normal screen and sacrifice the battery life.
Colour would also be needed to use the camera to take photos. At least, I would want that - and particularly once the automated processing is better.
Depends a lot on the usage.
Some people might want an eInk screen on the back side (for notifications and the current time and similar status items) and the existing screen on the front side. 
@Hristo i know you mentioned a few ideas but as a short list, what do you think would be your favorite top ~5 items that you would like Purism to focus on, on their next phone hardware development. These would be things that cannot easily or not be accomplished by software itself, and specs are important but aesthetics, usability, and availability could also be important factors?:
I can give you a top 1. 
In the current Librem 5 I miss most a second USB-C port.
Even better would be to also have a micro HDMI.
Why?
2 reasons:
If I had 2 USB-C ports, I could use one of them to charge the Librem 5 while using the other one for peripheric devices or a USB-C-USB-C capable monitor.
An extra micro HDMI would be also very practical as many monitors are not USB-C to USB-C capable. And I had a Nokia with a mini HDMI port almost 12 years ago!
This! I would like a port on top of the phone since that would be so much more convenient for my use case of a USB-C dock out the top then the phone itself on my desk facing me instead of upside down.
I wouldn’t mind a second USB-C port. I would probably class it as “nice to have” / “wishlist”, rather than high priority or essential.
However, as @Hristo says, this is strongly correlated with dock compatibility. If you find a dock that works and that has the right spec for your use then a second USB-C port becomes less important.
I think we can assume (?) that only one USB-C port will support video out.
From the description, I think both USB-C ports would need to support charging, which may be a challenge. And if you want to charge via both USB-C ports at the same time then that is trickier still.
You could tape to to the phone and treat it as a single bigger device
So the Librem 5 genuine Gnu+Lnx device still does not your dreams?? 
no no L5 is my dream except the screen not foldable 
I congratulate Purism for developing a free software phone using what appears to have been the only available means. But the Librem 5 current SoC (System on Chip) is extremely inefficient as a phone SoC. Purism or someone else needs to do silicon chip development to build a new cell-phone-specific SoC on a much smaller fabrication process, made specifically to meet Purism’s objectives. By putting all critical modules on one chip, the phone will burn far less current, thus allowing smaller batteries while giving much longer “on” and “stand-by” times. Such a chip could allow a Librem phone to be on par with any other modern phone as far as battery life and available features.
But silicon chip development is (can be) very expensive. However, if the opensource community did much of the development and used some foundry like TSMC to fab the chips, then the chips could be relatively affordable to consumers. Silicon fabrication only takes place in high volume. Either you fab at least one wafer with thousands of chips on it, or you don’t. However during the chip development process, most foundries put together shared wafers. So you (as a company like Purism) get a small part of one wafer (maybe only 100 chips), along side of chips from other silicon chip developers (usually big recognized chip manufactures) having their chips next to yours on the same wafer. For R&D, this works out well for everyone and is essential for the foundry to get new customers. For a consumer phone campaign, the risk would be higher. Maybe your first batch of 100 chips comes back with a fatal design flaw. But then again, maybe it comes back flawless. A handful of volunteer chip designers could build the SoC intellectually. Then Purism could take pre-orders for those who want their own chip from that wafer, along with the rest of the new phone model from Purism. Either that or the intellectual property is given to someone like NXP, to use freely under the terms of an opensource license, funded by another Librem campaign. But any way you look at it, the goal should be to get a dedicated modern phone SoC that is protected under an opensource license. Until that happens, everything has to be a compromise.
But if we ever do get such a successful opensource phone SoC in a Librem phone, then Todd’s dream will have been fulfilled. There won’t be any stopping a free software/hardware freight train at that point. Everything else being equal (including price and availability), why would anyone want a phone that spies on them, advertises to them, and that keeps them in that jail?
I’d settle for a processor die shrink from 28 nm down to 10-14 nm.
Does anything work like that? I recall that Google had some program about open source chip fabrication, but that was on some bigger process. Universities also etch their own chips, but I don’t know if they do that for general purpose silicon as opposed to specialized processes (e.g. optics).
Unless someone opens up a process library for small nodes, you can’t build open chips anyway. That was kind of the goal of the google project.
Meanwhile, I’m hoping for RISC-V competition to warm up.
I work for a large chip manufacturer. My job is to validate the functions of new silicon chips after the design work has been completed. So I am one of the first for my employer, to hook up the new prototypes when they come out of the fab, to see if they work as expected or not, and to see how well they work if they do work.
For chips that are fabricated internally, the chip manufacturing company just runs the new designs in one of their own fabs. But some chips are designed to be fabricated only by an outside foundary. In such cases, the foundary has to allow their new customers (chip manufacturers) to get small amounts of test chips made and tested, before their customer goes all-in financially, to get a large number of chips made. In small numbers, those chips are expensive. But they are available in small numbers. We once ordered and received only twenty of them. The price at that quantity certainly wouldn’t fit in to my personal budget.
The only reason I know about this is that I was the one tasked to pay for the chips and to run the initial silicon validation on them. It’s not like a publicly announced program. Not many people do this because the average customer for a founadry is a big chip manufacturer. But if you work for a large chip manufacturer that farms out chip fabrication work to foundarys, it’s very common to order small numbers of prototypes this way in the R&D phase of new product development. A part of the expense is that mask costs in the smallest of fabrication processes, can be hugely expensive. But costs come down when your mask costs and wafer space are shared with other chip manufacturers. It might be difficult or expensive (or both) for a small design house to get their new designs on to one of these shared wafers at a big foundry. You need a big foundry for the smallest fabrication processes. But it’s not impossible. It might be easier in a croud sourced program to order a whole wafer, after the initial prototypes pass validation.
Getting a good phone SoC chip designed and ready for mask making would be no easy task. A big foundary will have the critical design libraries available as a part of the services that they offer. If you use their libraries, your design seems to always work the first time. With new and/or less proven libraries, I’ve seen a fair amount of bad silicon (prototypes that fail to work correctly). But with a completed proven design, it’s just a matter of how to pay for production after that. But Purism hasn’t seemed to do well with big production runs either.
Another challenge for a modern phone SoC chip design would be the bringing together of many different technologies. Can the RISC‐V libraries handle the analog circuit design as well as the digital? Do all technologies used in the SoC, use the same libraries, made to run on the same manufacturing processes? Someone with more knowledge than myself would have to architect both the business and Engineering ends of an opensource phone SoC. But like the phone itself, it’ll happen because it has to happen if we want to continue living in a free world.
https://www.nasdaq.com/articles/an-overview-of-the-top-5-semiconductor-foundry-companies-2021-10-01
Librem 5 backpack?
