Thanks for sharing this. I'm one of those who once toyed with the idea of building a diffusion pump - allow me to share what I've learned in this quest:
The book "Creative glasblowing"(1968) by Hammerfahr and Stong have details how to build a two-stage mercury diffusion pump. You need to be skilled in working with boro-silicate glass (oxygen flame) to build this. I discussed this design with my local glassblower who told me his teacher passed away b4 he completed his apprenticeship, but the old master did tell him that the nozzles dimensions are very critical for the proper functioning of the pump. The book is out of print, I found a used copy via Amazon. It's a an excellent introduction to glass blowing, both scientific and artistic (my glasblower will not agree with this statement, but he's from the modern European school, the book describes the vintage American method using cross-fires)
I get the impression that with the exception of some Neon shops these designs are no longer used in industry. The only company that I could find that sell glass pumps are SVP http://www.svpneon.com/diffusion_pumps.html
. Of interest is the controversy surrounding the usefulness of these single-stage pumps as mentioned by Coyne on p.368 of "The laboratory companion" (1997). Blazek&Blazek in "Neon - the next generation"(2003) mention that they have found very few working diffusion pumps in the US Neon industry...
Soft glass do not need an oxygen flame, but I doubt if even a well anealed piece of soft glass will be able to handle thermal stresses (as mentioned by Geoffrey). And finding soft glass with the right dimensions might no be that easy.
In the "Proceedings of the A.S.G.S. 1995 Symposium" Wheeler describes how to build a 4" Glass Oil Diffusion Pump. Wheeler copied the dimensions of a commercial metal pump and achieved better performance than the original (375 l/s). I've attached some pictures out of the article, the dimensions look very similar to the Los Alamos design posted earlier by Steve. Once again you need to be well skilled in working with boro-silicate glass (at least a journeyman). Limited extra info at http://www.public.asu.edu/~aomdw/GLASS/DIFFUSION_PUMP.html
The manuals of metal diffusion pumps often contain detail of the internals, free downloads of manuals are available on both the Edwards and Varian websites. I've attached a picture I found in a Varian AX-65 pump manual. Looking at the drawings it should be easy to scale and copy using a small lathe (using Aluminium?). If you change the design finding the right heat input might need some experimentation.
As DIY'er the question I ended up with was if it would not be better and easier to build a Zeolite trap and not a diffusion pump? Published results suggest that you can equal the performance of diffusion pump with a Micromaze zeolite trap. I guess this refer to ultimate vacuum and that the pumping speed of trap will much slower? And unless I'm missing something the Micromaze is a commmercial simplification of what Kendall and David described in the article "High vacuum for teaching and research" in Am J.Phys 36(3),234, March 1968 (see tBJ summary http://www.belljar.net/82trap.pdf
). It's a relatively simple device with a heater to bake it out, the 13X zeolite is not that difficult to source. Not freely available is what's called 'Columbine formula 6' to bind the zeolite to the metal surface, but somewhere I read that you might be able to use water glass....
PS. Another example of (unfinished) homemade diffusion pump can be found at http://www.fusor.net/board/download_thread.php?site=fusor&bn=fusor_vacuum&thread=1172191668