I made more investigations and picked up all Olson timezones and their standard and dst offsets and if dst is observed, also the start time of dst and in dst-cases also hemisphere. The source was timezone-functions of php 5.3.5 and it's newest timezonedb version 2010.15. All dst observable timezones had dst start in 2011, excluding Fiji and Apia and it seems that they are not going to make dst-transition in 2011, so I take in these cases the dst start time of 2010 (september and october).
To find the most relevant or important timezone in cases when timezones were "unique", meaning all aspects (std offset, dst offset, dst start time) were the same, I made one more database search: the population of timezone. The source was database of 99 534 cities in [[http://download.geonames.org/export/dump/cities1000.zip|http://download.geonames.org/export/dump/cities1000.zip]] (all cities with population more than 1000). One of the columns in this database is the timezone of city. I think the population of timezone shall be the criteria to select the timezone. In detection script there are some timezones that could be replaced with more-populated ones, for example the following:
(See more on my home page, following the link below.)
These "unique" timezones are missing:
America/Sao_Paulo, Asia/Gaza, Australia/Lord_Howe, Pacific/Norfolk, Pacific/Kiritimati
The result of my research is in this address: [[http://www.kahkonen.com/loc/timezones_by_population2.php|http://www.kahkonen.com/loc/timezones_by_population2.php]]