Re: Ada.Calendar and NTP (and Unix Epoch)

["Vasiliy Molostov" <molostoff@gmail.com>]

Dmitry A. Kazakov <mailbox@dmitry-kazakov.de> писал(а) в своём письме Tue,  
24 Jul 2012 22:28:27 +0400:

> Furthermore, operations like UTC_Time_Offset, Time_Of, Split for some
> distant time in the past, e.g. 1901 are most likely wrong, because  
> OS/Ada's
> RTL cannot keep track of all political changes of all time zones in order
> to be able to evaluate the number of seconds from present time to 1901 in
> the time zone. It is *not* the number of years multiplied by the year's
> duration. (IANA Time Zone Database or alike is required for that)
>
> Calculating future times is absolutely impossible because that depends on
> future decisions of the corresponding governmental bodies having the
> authority over the time zone, from daylight saving time to changing the
> whole zone.

I took a look at  
http://www.ada-auth.org/cgi-bin/cvsweb.cgi/ai05s/ai05-0119-1.txt

And I can say that there is a mis .... "misviewving" of the Time, UTC time  
and Zoned Time (political)

While typical application takes time from some sort of monotonic time  
source (e.g. clock), which is supposed to be one and authoritative for  
such application, it is in the same moment this application also deals  
with time representation, and a "political" time (zoned) is one of such  
representations.

Probably and personally, I don't prefer to convert between representations  
and catch one or two hour diffs, instead I prefer to do calculations  
better done with monotonic time, and then convert result to the required  
representation.

Converting from some representation into authoritative monotonic time is  
"mission impossibe", since important data may be lost, because this  
important data belongs entirely to the representation, and can not be  
expressed in monotonic authoritative source.

Things like "ada.calendar is a fake" are just based on "speculative  
cone-moving" between "time real value" and "time representation".

>
> This is why it is better not to touch Ada.Calendar.Time.
>
> Ada.Real_Time, Duration, Time_Span, some numeric type would be much  
> better
> candidates.
>


-- 
Написано в почтовом клиенте браузера Opera: http://www.opera.com/mail/