We thank both referees for their helpful comments and detailed reading
of the manuscript.

Specifically with respect to the points raised by referee 1:

1) gfortran issues
------------------

We have extended the comments on the web page and the README to warn
about the possibility of gfortran giving incorrect answers as well as
crashing. 

2) Mass threshold and schemes
-----------------------------

Yes, of course eq.(6) only holds in the MSbar scheme. We've made this
clear in the text. We've also made it clear that we have not
incorporate mass thresholds in the DIS scheme at NLO, both 
on p.8 (now -> 9) and  p.26 (->28).

We've also removed the "nearly always" (residual heavy flavour when
crossing a threshold downwards) at the end of section 6.2.1.

3) Typos
--------
We've fixed the typos indicated by the referee, together with some
other ones that we have since found. We've also taken the opportunity
to add some missing reference and clarify one or two points that
seemed unclear to us.

A full list of changes can be obtained from

http://projects.hepforge.org/hoppet/trac/anydiff?new_path=%2Ftrunk%2Fdoc%2FHOPPET-v1-doc.tex&old_path=%2Ftrunk%2Fdoc%2FHOPPET-v1-doc.tex&new_rev=&old_rev=102


4) Sum rules
------------
The absence of tools for determining PDF moments has been rectified
with the addition of the TruncatedMoment(...) routine, now described
in section 4.4. 

An EvalPdfTable_Q(...) routine has also been added (described in
section 7.2) which returns a PDF object (i.e. a two-dimension array
with x and flavour dependence) for a given Q. Together with the
TruncatedMoment(...)  routine this provides means of accessing the
momentum and valence sums when using tabulations. An example of sum
rule verification is given also in
example_f90/tabulation_example_2.f90 .

As concerns automatic verification of sum rules: given the finite
range in x, one only has access to truncated moments and these do not
satisfy exact sum rules -- to carry out checks on these would involve
introducing parameters for tolerances, which would depend on the
x-range and on Q. We feel that the complications ensuing from these
parameters would outweigh the benefits of automatic checks on sum
rules.

