Can mdev replace udev in an ordinary linux distro?

[David Henderson]

On 09/20/2011 05:34 AM, Natanael Copa wrote:
> On Mon, Sep 19, 2011 at 5:44 PM, Laurent Bercot
> <ska-dietlibc at skarnet.org>  wrote:
>>> what worries me most is all the apps that depends on libudev (xorg,
>>> gvfs, etc). Those will still need udev for the full "experience"
>>   What these apps need is an implementation of the libudev API, not
>> necessarily the original libudev.
>>
>>   I have just taken a look at the libudev API documentation. It's not
>> overly complex, and it's sensible for the most part; I could probably
>> provide an alternative libudev fast enough if needed.
>>
>>   The real remaining problem is the library's backend, as in how to
>> design a simpler, more modular, but fully functional udev implementation.
>> mdev is a start, but it needs to be expanded if libudev functionality is
>> needed.
>>   I wouldn't worry too much, though. libudev-using programs are usually
>> behemoths that you would not run on embedded systems; on systems where you
>> need them, you can afford to run a full udev. Providing a full-fledged
>> alternative udev implementation is mostly aesthetics at this point.
> I thought the goal was to get rid of udev due to arrogant attitude
> from developer, doing things like requiring /usr merge in to / without
> caring what distros other than Fedora/RedHat thinks.
I agree.  But I think in order to do that, we'll need an mdevd binary 
that uses the libudev API as mentioned by Laurent.
>
> Its generally a good idea with standards and more than one
> implementation of a standard.
Again, I agree. If Laurent wouldn't mind contributing talents as was 
stated in one of the last posts, I think the mdev "suite" would be a 
very nice lightweight alternative to udev (with the advantage of not 
being dependent on /usr too)!  I know I'd use it as much as possible. :)
>
> If the goal was only to solve problem for embedded systems then would
> probably devtmpfs be good enough?
>
Now that Denys has applied the file reading patch, shouldn't that speed 
things up considerably for mdev?  And if so, what advantage would there 
be with using devtmpfs?

Dave