Accurate NMO-convolutional model by Symes. 
NMO and velocity analysis use Symes algorithm. 

linear operators that use the "addImage" function are not executing 
linear operations. Only the "image" function is a linear operation. 



We need to define clone and copy for Rfs (and maybe all other operators)
In the worst case we default it. 
look up meaning of clone and copy (I think clone=shallow copy (reference),
copy=deep copy). clone is now a standard in java. any author is responsible 
for creating copy constructors. 
To produce a standard "copy" of Rfs a user probably wants to clone its Space.


Naming: 
InterCapping ... capitalize letters within the word. Bad: hard to distinguish.
code should be case insensitive. letter cases should be used to make things 
more readable. A good rule: all new words capitalized within word (safes 
an underscore). begin of the word capitalized when a class but not a method 
or variable.

Product vector input for bilinear and semilinear: we definitely want it. 
derive bilinear and semilinear from operator so that general non-linear
solver will work. solver hierarchie is hard, I feel. bilinear op is true 
operator, etc. Hierarchy are entirely interfaces. we then implement only 
certain pieces that we are interested in. (e.g. a sepLinearOp implements 
a LinearOperatorWithAdjoint since SEP only codes linear ops with Adjoint.
sepLinearOp defaults the adjoint.image to an error message). 

How about changing the names "image" to "images" like in 
y = myNmoOpo.images(x)
(java calls it things "implements" "extends" etc) I love it! since it 
spells so nicely in the code and in the documentation. 
What are other functions that we should change (not get and set since 
they have to keep short. Is that inconsistency dangerous or stupid?)

We probabl want to reuse the example Bill and Mark showed last summer: 
Mobil Data set and an automatic velocity picking algorithm. 

Other great applications are: 
	statics (Wyoming data set of Rothman or newer data set. See 
		Dimitri's thesis)
	horizon fitting (Toldi and Arnaud)
	trace ballancing (Arnaud)




VERY DIFFERENT TOPIC: 

object orientation, geophysical data processing, a darwinistic approach, 
and some economical aspects: 

look at the various units of geophysical data: 
each data object should try to be the best object there is
	a trace tries to be as noisefree as possible 
		trace internat filtering.
		A trace may have to look at neighboring traces 
		to see what it means to be "good". But it competes 
		only locally. 
		maybe a trace should "invest" time in various activities 
		and make an investment in its goodness (assume risk)
	a cmp gather tries to select and weight its traces to make 
		the best gather it can. It buys traces that look good, 
		fit into its own concept. It again may have to look 
		neighboring traces on what it means to be good. 
	etc. 

	What should be the currency? (mean of exchange and trade .. 
		could be artifical and could be expressed in dollars
		as long as they are limited and bound to real resources). 
	What should be the selfish gene? (the underlying principle even 
		the various agents (class objects) may not have to be 
		aware of this overall selfish aspect ... they could at 
		more measurable effects such as dollars.
	What does such a processing world optimize? Maybe the most
		competitive place for processing?

	Failure of Tarantola: central processing scheme with one scheme 
		incorporating all processing steps. Optimize all of them 
		at once. Did he fail? Why did he fail? 

	Claerbout approach (industry standard): 
		optimize each little step but never think about the overall 
		processing. 

	New: combine them through market scheme. Does market already exist 
		but evaluation is done by people (processors) choosing 
		Claerbout's tools and toys while looking at the price tag. 
		Can that be automated. Can processes compete with each other 
		and die if unsuccessful. Can market only work when repeated 
		games (Axelroth: no collaboration in last steps). 



QWERTY: of course everything is localized to a certain degree: a great 
	seismic gather in the North Sea does not help you in Indonesia. 
	Local competition. 
