"Eric C. Weaver" wrote:

This discussion happens all the time on comp.dsp, between primarily
computer-science folks approaching DSP and EE types approaching it.

EE folks' definition of "linear" implicitly includes time invariance;


Interesting thought since a Signals and Systems course, or a Linear
Systems course, or a Communications course is often required to get an
EE degree. After all, these courses explicitly distinguish the linearity
property and the time-invariance property. And I've never seen the
"af(t) = f(at)" so-called "definition" until a few days ago.


DSP
people have to see it stated explicitly (as "LTI": Linear Time Invariant) lest
they think "linear" just means having no second-or-higher-order terms.

It is not a deficiency on either party's part, just a difference of definition
in each's respective discipline (is that enough alliteration?).

Therefore, I advise each to bend this much: Use the full phrase "Linear
Time-Invariant" when this miscommunication is suspected, so both know what the
hell the other is talking about. Now go and sin no more.

I've met folks before who think that linearity means freqs cumzoutas
must only equal freqs gozintas. But they don't usually put up such a
fuss when actually presented with the widely available and consistant
literature or reasonable arguments. This is more about fuss than
facts. Now that is consistant with the usenet.