[Shotwell] Ancient question

[Oleksij Rempel]

On 12.04.2012 06:38, bryan wrote:
> I've got images going back to the 1880s.  The program is
> mishandling older images (this is a known issue #3040).   Images
> appear to be disappearing (sorting incorrectly) within my 10k
> database and dating an image to say 1940 will also stuff up its
> event and today I just noticed a random sample of my images sitting
> in the rubbish bin (scary).  It's hard to pin down the date beyond
> which things go awry, for every test I do the limit seems to float
> around.
> 
> So...  does anyone know within what date range does Shotwell
> reliably operate?

This is interesting issue. The problem is in used time format. Usually
we use unix time (or m$ time:)). The reason is CPU usage.

So, i assume this kind of old photos are usually scans. And nobody
cared about EXIF data back in 1880's... This mean shotwell need to
create it. Best way to do it is XMP. We can use "xmp:CreateDate" to
store any kind of time according to ISO RFC 8601, with limitation to
year 0. For most historical collections it is OK. But i'm 100% sure,
if shotwell will make this step, after some time some body will ask to
support for some thing like 1.000.000 year BC :)
Any way. I thing the major problem is not to store the time in to
image, it is more about storing time in to database of shotwell.
In my historical study i know, that seconds and minutes in historical
collections and researches do not matter.

My suggestion to shotwell devs is to add timescale parameter to the
database. For example:
use same time base like before. start time is January 1, 1970.
scale = 100000 (seconds)
timestamp = 0000000010 (1970-01-01 00:00:10Z)
if we use set
scale = 10000 (minuts)
timestamp 0000000010 (1970-01-01 00:10)

scale = 1000 (minuts)
timestamp 0000000010 (1970-01-01 10)

scale = 100
timestamp 0000000010 (1970-01-10)

scale = 10
timestamp 0000000010 (1970-10)

scale = 1
timestamp 0000000010 (1980)

this kind of timescale use matroska for timestamping frames.

-
Regards,
Oleksij