Stack Exchange network consists of 183 Q&A communities including Stack Overflow, the largest, most trusted online community for developers to learn, share their knowledge, and build their careers.
Thank you very much for your answer (and also for your book of course :P). A'Campo's note is among the documents on my desk. Seems that I didn't miss anything. Then I think I'll start to look into it. For your information, the Coxeter groups that I'm particularly interested in are those of type (n-1,1) ("lorentzian" or "hyperbolic" in literature). We may have observed something interesting about their eigenvectors, but nothing is clear yet.
@David Cohen: Yes of course. But we need to assume that an embedding exists at the first place. I mean, situation like $G=C_3$ (triangle), $\ell=3,4,10$ should not be considered.