Reply To: Linux (debian) on cube i9

Reply To: Linux (debian) on cube i9

TechTablets Forums Cube Forums Cube i9 Forum Linux (debian) on cube i9 Reply To: Linux (debian) on cube i9

#58920
John Heenan
Participant
  • Posts: 5

Hi Carlos

Thanks for your reply. Your example of ease of us is very encouraging for others.

Antergos is exceptionally good for convenient installs on even recent hardware. For example if you want to just shrink a Windows partition and dual boot with Linux.

Antergos is built on Archlinux. The biggest hurdle with Archlinux is setting up the hard drive with the command line, however I believe it is common to use another tool for this.

I am not not confident there will be a fix that works ‘out of the box’ for wifi on good cheap tablets, like the Cube i9, in the near future, if at all.

I base this on the following:

The biggest market for Realtek wifi ICs is probably Android devices. Android devices do not use mainline Linux kernels.

Mainline kernels have very little practical use in day to day cheap devices and the mainline kernel maintainers are divorced from reality when it comes to issues that interfere with their pet projects, evangelical zeal and heroic self portrayal. There is almost a pathological avoidance of real world tests when it comes to network devices as it would probably just nail more coffins in the mac80211 highly abstracted driver they think they can convert manufacturers to.

While my own tests revealed I could get a mac80211 based driver to work for the Cube i9, with modification, the performance was abysmal and there was an awful lot missing. As far as I am concerned mac80211 is nothing more than an idea that Linux maintainers have failed to get manufacturers to get behind. Further tests have indicated it would more than likely take an enormous amount of work to ‘get right’.

In more senior Linux levels, there is now almost a pathological revulsion against drivers with very high abstraction layers. For example the recent attempt by AMD to have very high abstracted video drivers for forthcoming hardware accepted into the mainline kernel was flatly rejected with a ‘no, not ever’ attitude.

So it looks like there is internal division within Linux towards unnecessarily high abstracted drivers with:

1) The more experienced and senior hostile towards them.
2) Advocates averse or hostile to what is beneath them: real world tests (the great destroyer of vision), instead probably blaming others for lack of vision.

3) Manufacturers perplexed: “We have done the right thing at enormous expense to get amazingly high performance with long established and accepted techniques and no tricks. Now you are telling us we have not done the right thing and expect us to use tricks dictated by you for a tiny specialised Linux market that has no real interest in cheap mass produced ICs, unlike Android device OEMs. Who do you think you are? Get stuffed.”

So where does that leave us? An AUR package (usable by Antergos) is one way to go to make it easier with kernel upgrades. However the current requirement to downgrade networkmanager does not look good. Also it still leaves a nasty problem in place: initial installation from just an install image that requires networking.

There will be cheap more capable 64bit ARM SOCs with integrated wifi which may alter the landscape. Windows will use them to emulate x68 applications.

John Heenan

Lost Password

Skip to toolbar