Title: "spectral" [atlas] . "spectral" usually refers to real spectra,
or the reader may think that's what they find in the paper.   
"broad-band SED"  more appropriate ?

--I added "broadband" to the title.

Tot mags have been derived in a slightly different way for UV than for
optical/IRAC/MIPS.  It would be very useful to quote in the tables the
radii used (and ellipt. and PA) for comparison to other past and
future measurements.   You mention that UV asympt. fluxes
are on avg 14% larger than UV fluxes at opt. radius: is this
the opt. radius you use for your opt. mags?  How about quote both 
in the tables: i.e. I suggest to add in the tables: adopted opt. radius
(used in yr meas.), [ tot. area ], and UV flux both in this
area and asympt. value.  Also, it would be useful to add adopted 
distances, type etc. in a first table (or beginning of table 1) 
and you could also put the ext. there (i.e. all data that are not 
your measurements or derived in this paper given first).

--I have added a Table 1 that lists optical morphologies, coordinates, and
  the apertures used in optical/IR flux extractions.
  I don't use distances in this paper, so I'd prefer not to jump into
  the distance game here.
--Armando has re-extracted UV fluxes using the same apertures that I 
  used for the optical and infrared flux extractions.  He finds very
  few differences in the two approaches.

UV total mag: were the foreground stars removed from the UV
meas. as well ? (in the same way as for the global optical fluxes?)

--The UV magnitudes were measured after properly masking all field stars
  and background galaxies.

"infrared-to-uv"  define here that "infrared" is total ir , and why.   
This is my major question actually: if you use total ir , you include 
dust (--> young) and old stars, and these two components I'd expect have
a different relationship to UV.  We can discuss more about that 
off-line, to keep this short.  Anyway, better define what yr assumptions
are - although it's obvious from the figure captions,  I think it's nice
to elaborate in the text

-- Total infrared here refers to only the dust continuum emission between
   3 and 1100um.  I've added this clarification to the text.

"broad dispersion" - perhaps better to call it "spread" than dispersion,
to indicate that it is a real range...

-- Done

Sect. 5.4 That's because the UV 
slope is much more sensitive to age (and to the extinction which is 
extremely different only in cases of strong starbursts [from Calzetti
papers ]) and in particular the intrunsic UV slope changes when 
populations older than 100Myr dominate of significantly contribute.

--Words to this effect have been added to the text.

4.3 - I don't understand how these relate to interpretation
(but maybe that's coming later ? )

--Good point.  I have added a connection between the eigenspectra
  uncovered by the PCA and the additional dispersion that early-type
  galaxies give to the IR/UV vs UV slope plot.


typos and small things:

-- Fixed.  Thanks.
=================================================================
 I read your new version and I don't have new substantial
comments. (some typos and small things that you'll find anyway).
 Only thing - I wonder about your WEDGE in fig. 13 : could
it be that the lower left corner emptiness is due to
selection effect / incompleteness - and the real thing
is not a wedge but just the upper "envelope" ?
(i.e. the upper left corner emptiness is real, the ll
corner maybe a bias ?)

--Grazie, Luciana.  I've thought about the same question, and I've 
added a few words on this.  I've added data from the GALEX Atlas
of Nearby Galaxies, and they show a very similar distribution.