Since the publication (Karsai et al., 2002) we have improved the home-made SYBR-green mixture further:

Our new formular for the 10x buffer is:
100 mM Tris-HCl, pH 8.5, 
500 mM KCl,
20-30 mM MgCl2 
1.5% Triton X-100
Please omit completely the BSA. 

It sounds funny but it is better not to make more than 2 ml aliquots of the 10x buffer because yet unidentified reasons some mixtures do not work. However if an aliquot is working it can be stored forever.
 
Concerning the MgCl2 concentration. Start with 2 mM (final concentration) and if amplifications are bad rise the final concentration to 2.5mM and 3 mM.
 
Our pipetting scheme:
Prepare Mastermix of
     water
     10x buffer
     primers
     SYBR green
     TAQ
 mix well than start with the pipetting of the reaction
1. template
2. aliquot of the mastermix.

The mastermix is stable for about half a day. Thus, if two runs with similar mixtures are planned - prepare one common mastermix. Do not freeze it and do not store it over night.
 
Another very important issue is the stability of the SYBR green:
We bought it from Roche since it was stable for more than three years.

Concerning the storage: Aliquot the concentrate in about 50x  10 µl portions and store them at -20°C in the dark (i.e.aluminium foil protected). These aliquots can be thawed about 5 times.
 
Aliquots are diltued by 1:1000 with TE pH 7.5 (sometimes even water works - but this depends on the pH - as low as pH 7 works) as described in the paper. This diluted SYBR green has to be use within one week (maximum two weeks).

Please DO NOT thaw/freezing the 1:1000 dilution too often. Actually it is better to store it protected from light at 4 °C.
 
Recently I go following email from a happy user:
"I got really good results with your precise advices for the homemade SYBR green. Now our homemade reaction mixture works at least as well as the commercial kit. You have to know that the SYBR Green dye is more stable (we use the same tube since the beginning of this project) and efficient when diluted in DMSO (dilution 1/100) rather than TE. We also add 2.5 % of formamide by reaction to improve the specificity."