Not so much EP related, just maybe the idea: ZX Spectrum related maybe something similar "multiple purpose" solution: https://sites.google.com/site/interface1bis/

Anyway this is too modern for my taste I like the things what I also can build.

But need to design same with the existing one, or agree with the developers.

Indeed.

But need a terrible surface mounted components which can't handle at home with any user

It's a question in general, what kind of really good all-in-one solution can be created with avoiding SMD parts, programmable logic etc. So all the things which are hard for an average hobbyist maybe.

(Ethernet) I know someone working on it

Yes, but the problem is more like "too much needs and too many of discrete projects", it's not so easy to connect millions of extensions, maybe the "all-in-one" solution can combine many of these, that was the reason I mentioned.

I think the USB too complicated thing with lot of protocols need to be implemented. Then will be slow compared with direct SD or CF cards. I did not feel it necessary to.

It depends. If you need to create it from discrete parts, only logic ICs, it would be quire hard However if some MCU is involved in the design, it can be not so hard, also what I mentioned already there are even USB controller specific ICs which help a lot for the user to avoid the complexity to do the bulk of the USB related issues by themselves. Maybe CH375 or similar IC seemed quite handy here, but now I can find only web pages in Chinese (indeed it is Chinese, but for an English specification at least), it seems the English page is removed and/or replaced with Chinese one now, ouch.

Also, please note, that it was more like a "dream wish list" not so much a realistic one in every details

You can't done as compatible on external card. Need to connect to joy port, or inside the machine.

True, but I also thought as an internal all-in-one solution not only external solution, thus mentioning keyboard and mouse too.

Also possible inside the machine. But why? Todays are available new membranes.

Maybe I am just too much lazy again and used to PC keyboards, even my old favourite (C64 keyboard) feels too "alien" for me nowdays compared to the daily used PC keyboards, especially better ones. Well, that's my problem, yes

I'm love floppies! The tak-tak-tak... needed for the retro feeling!

Probably, personally I still hate floppies whatever retro or not, but again this is kinda personal point here

UPDATE:

Found the USB stuff: http://www.8bit.8u.cz/Files/CH375/CH375DS1.pdf

Ok, it's still not DIP, but 28 pin SOP package is till (?) better than a 100000 pin monster QFP or even BGA It seems it has parallel bus interface, and command set to implement storage protocol without too much actual details but honestly I am dumb with USB just a quick peek on the document what I've done ...

Today, collect materials for card design:
- 512 KB Flash ROM
- up to 2 MB SRAM (4x 512K x 8bit)
- RTC

I found the original parts available MC146818, but if there are other newer and more available, I will use them willingly.
Is there a list of counterparts, do I need to do some research yourself ?

IDE card design is actually ready, but I have changed a little idea of FlexiBridge (I want to be more compatible with the classical single bridge).
One channel (master/slave) with CF slot.
512KB flash on memory card is enough?

@gflorez:
No satanizar, por favor, mi trabajo Estoy empezando a preocuparse por la Inquisición

Of course, I'm joking (I hope that the translator did not spoil the sense of what I mean).
A IDE module is not built in one and a half day, but I drew it for about a week.
On average, one project will take me about 1-2 weeks time.
From the beginning to collect information about the interface, then the availability of components, the prices (at once at this stage).
Then gather notes directory components, usually worked up a bit of the original design and draw a diagram.
Designing the PCB takes about 3-5 days, depending on the complexity and density of the circuit.

In the case of IDE nothing I transformed, and all components are still available, so it was faster.
And today I'll draw a diagram RFC card (RAM-Flash-real time Clock).

@all:
Most cards have a data bus connected to a computer through the buffer 245.
Should I separate the memory data bus by buffer ?
It seems to me that, yes, but I have doubts about the same buffer.
Maybe I will apply a series of ACT (fast and energy-efficient).

As a friend of mine says: "I'm done 40 years, I can not waste time"

Should I separate the memory data bus by buffer ?
It seems to me that, yes, but I have doubts about the same buffer.

The problem, especially if you allocate all address space
Without buffer, then just left empty some sockets if needed to cooperate with other existing expansion.
With buffer it is still send the FFh to the bus when the socket is empty and make a collosion on the data bus. At this mode need a enable jumpers for each socket which is can disable the bus drive and the chip enable (then not needed really remove the chip).

Anyway I really don't know more ROM programs than 512K I suggest the secondary ROM socket with jumpers for FLASH/SRAM mode.

Other suggestions which is make the address decoding more complex

For using lot of memory needed my Quick Memory test. One reason the original EXOS test testing megabytes of memory for minutes at each reset... will be better done at few seconds
Another reason: the original EXOS only find ROM extensions at x0h segments and in the cartridge (04-07h). My test can find ROMs at any segment. (And also possible simulate ROMs in RAM).
The best way is using my unofficial upgarded EXOS versions (then some EXOS bugs also fixed), but need to dismantle the machine, and modifiy the ROM socket for 64K.
Other way use the cartridge version. This is needed to placed to 04h for the EXOS can find it.

Then will be good idea if 64K of the Flash ROM can be moved to 04-07H (cartridge area) with a jumper settings. (for example the 24-27h moved) Then possibel to run cartridge version of the Quick Test with untouched machine and without additional cartridge.

Other suggestion for EP64: unfortunatelly most of EP128 programs don't allocate the memory from the EXOS using direct segment address from F8h-
Then needed to put 64K memory expansion to F8-FBh for make EP64 compatible with EP128. Then the suggestion are EP64 jumper which is move 64K of one SRAM to this area.

Maybe I will apply a series of ACT (fast and energy-efficient).

It is good idea! (I using HCTs from early '90s)