In my opinion, analysis in this context is computing and
    attempting to differentiate leads to a slippery slope, so if
    somebody wanted to sidestep the rules then they just declare their
    computation is "analysis".  For the high-throughput biological
    research I mentioned before, that "analysis" is critical to the
    interpretation and understanding of the physical measurements. So
    while such analysis may be ancillary, it really is required for
    extracting biological meaning from the data.

You may be right.  I had better think further about this.

The difficulty is that there are always small computing activities
associated with doing anything with data.  To say that those small
activities make any service that does things with data into SaaS seems
like going too far.  So it is a matter of where to draw the line,
and I am not sure there has to be one right place, or where it would
be if it exists.

    Regardless this still leaves the issue that significant
    computation, e.g. assembling millions of sequence reads into a
    genome, is beyond the capacity of the computers in a typical
    scientific laboratory, so investigators rely upon SaaS to do their
    research.

I wonder, is it really SaaS?  Using a remote supercomputer
is not necessarily SaaS.  It depends who controls the software
that does the work, and on other aspects of the situation.

Even if it is SaaS the way it's done now, it doesn't
inevitably have to be SaaS.

--
Dr Richard Stallman
President, Free Software Foundation
51 Franklin St
Boston MA 02110
USA
www.fsf.org  www.gnu.org
Skype: No way! That's nonfree (freedom-denying) software.
  Use free telephony http://directory.fsf.org/category/tel/