On 4 May 2005 at 10:22, Jim Starkey wrote:

>OK if I do them as timestamps? Since dates are stored as GMT/Zulu and
>we're GMT+5, the first one isn't strictly what you want; we'll use
>31-DEC-1969 18:59:59 instead. The one second or 1,000,000 nanoseconds
>before the base date, so the physical value, an int64, is -1,000,000 or,
>in hex FFF...F0BDC0. The encoding is:
>
> edsNanoSecLen3, 0xF0, 0xBD, 0xC0.

Then, encoding procedure looks too complicated for me. It must
strip leading empty or full bytes and check that first valuable byte
after a full one has sign bit set otherwise - keep the last full
byte.
Does it worth efforts? If I understand you correctly, encoding-
decoding must happen every time data cross module boundary. I.e. from
engine to client application it may occure several times if providers
are stacked. How many encodings will occure on this way: engine
provider -> Y-valve -> Network provider -> Network provider ->
Y-valve -> User application?
--
SY, Dimitry Sibiryakov.