> > sentence and word have some meaning within a MARC context set, but 
> > paragraph doesn't really.
> 
> Sure it does.
> 
> 500$a This is a paragraph note with two and three sentences. 
> This is the second sentence of the first paragraph\n\nThis is 
> the second paragraph and third sentence.

Apart from the fact that in MARC \n means afaik err \n not newline. In
fact, afaik there isn't a standard encoding for newline in a MARC field.
 
> > Mainly you don't normally associate sentences with MARC records 
> > (especially if searching in indicators!)
> 
> Sure.  I wouldn't normally do it either, but someone asked 
> what it meant :)

Yes - and others will - not valid within MARC is a perfectly acceptable
answer.

> > marc.500:1 any 1 prox/unit=element/distance=2 marc.500a any baz
> > 
> > mean? 
> 
> Exactly equal to two elements apart.  The above query is not 
> meaningful, however given a different pre-supposed structure 
> it could be.

Ditto - someone on seeing distance=0 will ask what happens if
distance=1, etc. Not valid is an acceptable answer.

In fact these aren't hypothetical questions - any server developer will
need to know how to react to MARC searches that come in with unit= and
distance= set to other values that element and 0 respectively.
 
> > Well then we have a problem - your last example works 
> because 500$a is 
> > included within 500.
> > 
> > However, in the case of
> > marc.500:1 any 1 prox// marc.500a any baz
> 
> Correct.  That's not a 'problem', it's an empty result set.

Well, it is a 'problem' in that the above was originally *your*
suggested "cleaner" encoding (than my relational modifier approach) for
queries of the form

Records containing MARC 500 field with an indicator1 value of 1 and a $a
subfield which contains baz

Matthew