Simon Bridge said:
I think Purcell is boring myself ... probably just the writing style conflicting with the way I read.
I don't think it's boring. It is one of the early attempts to teach electromagnetism in a modern way and not to confuse students with the usual non-relativsitic treatment of the sources and then making up puzzling paradoxa like "hidden momentum" that in fact were solved very quickly after Einstein discovered special relativity. I think at latest in 1912 von Laue came up with the correct solution to all these problems. Nevertheless it's still taught the old pre-relativistic way in even some newer textbooks. Only at the very end you find a brief chapter on "relativistic electrodynamics", which in fact is a tautology, because electrodynamics is a relativistic field theory since Maxwell discovered it, although this was hidden for almost half a century, and it was Einstein's real breakthrough to figure out that one has to change the spacetime model affecting entire physics, including the "mechanical part" and not only electromagnetism. Thus that's how electromagnetism should be taught since 1905.
However, the book by Purcell is overly complicated and sometimes confusing, as the long debates we had about it some time ago in these forums. In my opinion one should start right away with the mathematical description of special relativistic spacetime in terms of the covariant formalism a la Minkowski (of course not with the confusing introduction of imaginary times but as a real pseudo-Euclidean affine space; imaginary times are needed only in quantum statistics and they have nothing to do with relativity) and then use simple relativistic models for matter, i.e., charge-current distributions, to present electromagnetism in a manifestly covariant way.
The SI is a pest in electromagnetism. Why should electric and magnetic components of the one and only electromagnetic field be measured in different units? Of course, sometimes we are used to that. In the US distances on the road are measured in miles, and the height of mountains or bridges in feet and inches, but that seems not to be too attractive. You can even argue, why to measure distances and times in different units and not setting ##c=1## right from the beginning. I don't think that latter idea is a good one, because indeed spatical and temporal directions are dinstinguishable in Minkowski space due to the pseudo-metric's signature (1,3) (or (3,1) for the east-coast convention) arising from the necessity of establishing a causality structure of spacetime, and times are qualitatively different than spatial distances. So there's no argument for measuring ##\vec{E}## in different units than ##\vec{B}## but some justification for measuring times in a different unit than dinstances. So I think the best choice of units is the modern rationalized Gaussian system, the Heaviside-Lorentz systems, which is used in high-energy particle and nuclear physics.