In designing primers for PCR, the following steps/rules were tested and proven to be useful:

length of individual primers between 18-24 bases. Longer primers (30-35 bp) seem to work in more similar cycling conditions compared with shorter primers, and can make multiplexing easier (see picuters below).
it is desirable (but not absolutely necessary) that the two primers have a close melting temperature or Tm (say, within 5o C or so). If Tm difference between the two primers is high, the lower Tm can be increased by increasing the length of that primer at the 3' end (this can also keep the size of the amplified locus constant) or the 5' end.
purine:pyrimidine content around 1:1 (maybe 40-60%)
if possible, primer sequence should start and end with 1-2 GC pairs
each primer pair should be tested for primer-primer interactions. For this purpose a useful Macintosh program is "CPrimer", a freeware available at ftp.bio.indiana.edu. This program also provides the melting temperature for the sequences entered, thus helping in designing PCR programs. Very convenient, some web sites offer programs that can be used directly on those sites to do the same functions: (search for optimal primers, melting temperatures).
primer sequences should be aligned with all DNA sequences entered in the databases (using BLAST programs) and checked for similarities with repetitive sequences or with other loci, elwhere in the genome. If two loci are very similar (for example across species) it is useful to design the primers so that at least 1-2 bases at the 3' end are specific for the locus to be amplified
cycling conditions and buffer concentrations should be adjusted for each primer pair, so that amplification of the desired locus is specific, with no secondary products (see other pages). If this is not possible, the sequences of the primers should be either elongated with 4-5 bases or simply, changed entirely.

	Fig. 7. Multiplex PCR using primers 18-24 bp long When PCR reaction Eight individual loci are amplified with similar intensities when the primer pairs are used separately. When equimolar amount of these primers are mixed together for a multiplex reaction (Mix K), some of the products are much weaker (#1, #2, #5, #6) than other. In this case, primers had "usual" length, between 18-24bp. (primers used in this case amplify polymorphic loci, explaining the "double" or "triple" bands as seen on a regular agarose gel)
	Fig. 8. Multiplex PCR using primers 30-35 bp long Compared to the figure above, in this case the primers used for multiplexing were longer than 30 bp (up to 37 bp). Equimolar amounts of primer were used and all loci were amplified with comparable intensities in each reaction.