Trent Piepho [off-list ref] writes:

The signal recevied by the ir receiver contains glitches.  Depending on the
receiver there can be quite a few.  It is also not trivial to turn the raw
signal sent by the remote into a digital value, even if you know what to
expect.  It takes digital signal processing techniques to turn the messy
sequence of inaccurate mark and space lengths into a best guess at what
digital code the remote sent.

This is of course true. Except that most receivers do that in hardware,
the receiver/demodular chip such as TSOP1838 does it.
If you receive with a phototransistor or a photodiode feeding some sort
of ADC device (not a very smart design), sure - you have to do this
yourself.

I have never heard of such receiver, though.

One thing that could be done, unless it has changed much since I wrote it
10+ years ago, is to take the mark/space protocol the ir device uses and sent
that data to lircd via the input layer.  It would be less efficient, but
would avoid another kernel interface.  Of course the input layer to lircd
interface would be somewhat different than other input devices, so
it's not entirely correct to say another interface is avoided.

IOW, it would be worse, wouldn't it?
-- 
Krzysztof Halasa