Hello!

I am trying to learn more about timestamping. From the definition of
timestamping as given on major sites, it appears that it was designed
to vouch for the existence of a document at a specific time and to
track different versions of a document (by differentiating it with a
timestamp).

However, I wanted to find out if you were given a timestamp from a
server cluster or a distributed database, could it be used
(absolutely) to point back to a single file? i.e. would it be
absolutely unique?

If the servers or server clusters are powerful enough with multiple
disk drives to write data, would it not be possible for several
documents to be saved/written at the same time (however remote that
chance may be). I am looking for arguments that support this line of
thought, possibly with some references to other documents/research
papers or technical comments that explain the same.

If you need any clarification, please feel free to contact me.

Pat79-ga

This is what I know of Timestamping and disk clusters:

Depending on the file system, clusters are generally in a
semi-orginized format. Without a MFT (Master File Table) it is
possable to extract data from a disk drive by scanning every sector.
On many file systems the MFT only tells the program where a file can
be found on a drive and how big it is, from there the program reads
each sector or cluster until the specified size is reached. However,
(depending on the file system) the MFT sometimes contains the info for
the timestamp, but often stored in the info at the begining of the
file (where the name, [often a repeat] size, and [often a repeat]
timestamp for created, modified, and accessed is stored).

As for pointing to a file from a time stamp, it is possable, but
unless the file system or database is made spacificly for to do that,
it would not be very fast, seen how it would have to search each
sector (and possably the MFT).

You may also want to know that it is impossable for a hard drive to
write more than 1 thing to the drive at a given time. The head of the
drive can only be either reading, writing, or seekig info from the
drive at a given moment. It would be possable however for a hard drive
to be made with 2 or more heads on it, in that case it would be
possable to have both heads writing data at the same time. Also, a
computer can only do 1 process per processor at a time; so if the
current process told the drive to write data to the disk and the next
process was also telling it to write data to the disk, if the drive
was already ready for another command, there is NO way to fill that
gap without using some sort of multi-process machine (but this
mostlikely would never happen seen how drives are almost always much
much slower than the processor).

However it would be possable to make a cluster of drives and make a
specialy designed progam that would write and read data from each
drive seperatly, then compile the data for the output or input. But to
get the program to work correct, it would have to space each bit of
info evenly thought each drive. This would dramatically increase
seeking time for a drive, but would require more processing power.

To get back on topic...

To make a new file system or database structure that uses timestamp to
place a file may not be the best idea, unless you reserved a place for
the occurance of more than 1 file being made at the same millisecond
(which may happen often, especially when installing programs). This
may not be a great idea seen how in January 19, 2038 the Unix time
stamp will swich to a 64 bit number and cause another big  "y2k"
similer problem.

I rambled on a bit in there, but that is much of what I know about
timestamps, and the way computer reads and writes data.

Hope that helps,
-ALL