A few months ago I was introduced to the new DateTimeOffset type and was glad DateTime\'s flaws with regard to time zones were finally taken care of.
However, I was
Well, one obvious answer would be when you need to support clients without the SP that it ships in (it isn't actually in 3.5 - it is in 2.0 SP1, which shipped at the same time).
Sometimes you really just want to represent a "local" (timezone unaware) date and time rather than an instant in time. To be honest it's more often useful to represent just a time - e.g. "wake me up at 8am, regardless of timezone" - but date and time could be useful too.
I agree that for the vast majority of cases, DateTimeOffset
is a better fit. It does strike me as odd that there isn't a DateTimeTimeZone
struct which has both the instant and its timezone though... an offset doesn't actually give you all the information you need. (For instance, given a DateTimeOffset
, you don't know what the time will be 24 hours later, because you don't know when DST might kick in.)
If you want that kind of structure, I have a very crude implementation in another answer. I'm sure it could be improved very easily :)
Whilst I wouldn't PREFER to use DateTime over DateTimeOffset, please note that sometimes you NEED to, as MS .Net does not support DateTimeOffset as a DataColumn.DataType property DataColumn.DataType Property even though SQL datetimeoffset has been around since SQL2008.
I myself had a problem with reading (ReadXml
) a DateTimeOffset value of an XML exported DataSet with XmlReadMode.InferTypedSchema
; it read it as a DateTime and crashed when I tried to merge it into a DateTimeOffset column