The lack of spreadsheet functions was just an oversight. I made the original list back in 2012 and probably just missed them.

gsvd has moved to core Octave dev version.
All that is left now is funm.
But in the long list in comment #0 the spreadsheet I/O functions weren't mentioned. JWE once wrote he wanted to move them into core as well. I've been working on it and the last thing I want to do before switching is remove duplicate code for ods... and xls... functions; most of that is done already in my private repo at home.
Plus some discussion may be needed on the maintainers ML on dependence on Java + external support SW, or optional, or no dependence at all, and in the first two cases, what Java support SW.

Now that ODE solvers are in core I think also
ode23s (odepkg)
should be added to the list

Wasn't there a GSOC project or the like for a better funm? It has been advertised for several years on the projects idea page, but IIRC this year there was some movement.
Anyway, the funm in linearalgebra and its intended replacement from Higham's site differ from contemporary Matlab funm.
AFAICS the latter nowadays has different (stricter) requirements for the function to apply to the matrix argument and accepts an options struct as optional 3rd input argument.

I moved condeig from linearalgebra to the core in this cset (http://hg.savannah.gnu.org/hgweb/octave/rev/e859caa53399).
List of functions remaining to move (7/18/2016):
funm (linearalgebra)
gsvd (linearalgebra)

The function publish has been moved to core (see patch #9048).
So the list of functions needed to be moved is now:
condeig (linearalgebra)
funm (linearalgebra)
gsvd (linearalgebra)

Much progress has been made.
The current list of functions (9/3/2013) is:
condeig (linearalgebra)
funm (linearalgebra)
gsvd (linearalgebra)
publish (miscellaneous)

I added a note in the unimplemented.m script pointing users of exifread to imfinfo (http://hg.savannah.gnu.org/hgweb/octave/rev/84c119957463).

I was looking closer at Matlab's does and in addition to the minor list I mentioned before, they also returned 2 structs with the rest of the Exif data.
Anyway, I have implemented this in imfinfo() and deprecated exifread in the image package. This function can be removed from the list.

I like option 3. This preserves Matlab compatibility, but also allows users to get as much information as possible about the image. This fits the Octave ideal of being able to run any Matlab script, but also extending syntax where it makes sense.

Should we bother wth exifread? Matlab has deprecated it. They suggest to use imfinfo and I looked into it, but matlab's imfinfo only reads a very small part of the EXIF data, in no way comparable to what exifread/readexif does.
I have 3 suggestions for this:
1  make imfinfo read only the same EXIF fields has matlab;
2  make imfinfo return all the EXIF fields that exifread/readexif at the root of the structure returned;
3  make iminfo return the same as matlab with only one extra field, named EXIF which itself is another structure with all the other EXIF fields.
What do you guys think is best?

The list of missing functions is shrinking.
New list as of 5/3/2013
condeig (linearalgebra)
exifread < readexif (image)
funm (linearalgebra)
gsvd (linearalgebra)
publish (miscellaneous)

Nice. I have removed them from the specfun package.
By the way, the ellipj function which is implemented in C++, used to also have a .m implementation as well. It performed a bit slower (not by much if I remember correctly), but may be possible to improve.
Since m file implementations are usually prefered, the change occured with r9384 in the specfun package
https://sourceforge.net/p/octave/code/9384

The specfun functions ellipj, ellipke, and expint are now imported from Octave Forge. I've also removed resample from the unimplemented list.

New list as of 3/14/2013
condeig (linearalgebra)
ellipj (specfun)
ellipke (specfun)
exifread < readexif (image)
expint (specfun)
funm (linearalgebra)
gsvd (linearalgebra)
publish (miscellaneous)

Please note that the Matlab core function resample is actually a method of the timeseries class, at least this is what the API docs on the web tell me:
http://www.mathworks.com/help/matlab/ref/timeseries.resample.html
The resample function that operates on standard data types is in the signal toolbox:
http://www.mathworks.com/help/signal/ref/resample.html
If someone can confirm, I think we should remove resample from the list of unimplemented functions to avoid confusion.

I'd also like to have
ode45 (odepkg)
ported. It's in matlab core and often requested by octave users.
c.

Progress has been made. The list of functions to be ported has narrowed. The current list (8/14/2012) is now:
condeig (linearalgebra)
ellipj (specfun)
ellipke (specfun)
exifread < readexif (image)
expint (specfun)
funm (linearalgebra)
gsvd (linearalgebra)
publish (miscellaneous)
resample (signal)

Octave has a list of core Matlab functions (unimplemented.m) which are missing from core Octave. Some of these functions are already available in OctaveForge in various packages. It would be a good idea to bring these functions into core Octave and update their coding styles to reflect core Octave. This is a much less arduous task than trying to create these functions from scratch.
The following functions with their package in parentheses have been identified:
cmpermute (image)
cmunique (image)
condeig (linearalgebra)
ellipj (specfun)
ellipke (specfun)
exifread < readexif (image)
expint (specfun)
fminsearch (optim)
funm (linearalgebra)
gsvd (linearalgebra)
publish (miscellaneous)
resample (signal)
sound (audio)
soundsc (audio)
