unfortunately it has been found, as far as I know, that the joint-maximization (Shall,1991) and the penalized quasi-likelihood (Breslow & Clayton,1993) methods in general performs poorly (take a look at Engel B., Buist W., BIOMETRICS 54: (3) 1155-1164 SEP 1998 and McCULLOCH C.E., JASA ?? 162-170 MARCH 1997). The Monte-Carlo Newton-Raphson method suggested by C.E. McCULLOCH (1997) seems to be a good alternative, but I don't know if it can be applied to the animal model. Another useful reference is Tempelman, May 1998 Generalized Linear Mixed Models in Dairy Cattle Breeding, J. Dairy Sci. 81:1428-1444.
Asreml mailinglist archive: http://www.chiswick.anprod.csiro.au/lists/asreml