[KPhotoAlbum] md5sum calculations

[jedd]

 Hi Eivind,

 Thanks for the clarification.  I've done some large-scale relocation
 of directory structures around, but I always do a full re-calc checksum
 operation prior to that - shut down KPA , do the move, and then
 start KPA and ensure tags survived the move.  Oh, and of course
 they always do.  :)

 When you say md5sums are recalculated when the path or filename
 changes, assuming no internal changes to the file, then the md5sum
 will be exactly the same as it was before, right?  This is my
 understanding of md5's - that they are purely content based and
 not on any file meta data - but just want to make sure that KPA's
 doing what I assume it's doing.

> So, renaming files *will* cause kphotoalbum to do md5sum
> recalculations for the moved files. If you recalculate md5sums (from
> the menu in kphotoalbum) and thereafter save, this problem should go
> away. (until the next time you rename files, after which it
> reappears)

 I think what's happened is that I've done a few scripts to fix up
 orientation and date-time stamps on my files (exiffile -t to make
 the file's date match the exif date).  The latter shouldn't make
 the md5 change - it's external meta data to the content - but I'm
 now thinking that the jpegexiforient utility I'm using makes a change
 even if the orientation is already set correctly .. hence the md5sum
 gets re-calculated and changes between these.

 I have two problems here (neither of them are KPA's):
 o  I'm trying to sync my large (50GB) stash of pictures from my
     laptop in London to my server in Sydney, periodically
 o  my web front end I'm playing with uses a vaguely RESTful
     approach, and the UI assumes some consistency with the
     md5sum's, insofar as handing URL's around to people.  

 I'm not sure how I'll deal with the second option - other than to
 be more careful and/or run my jpeg-exif-orientation stuff over
 all my pictures really soon, and then sit on my hands.

 The former option I'm hoping will go away with a combination of
 running the same orientation fixing scripts at the other end, and
 then relying on the smartness of rsync or unison to sort it out
 with minimal bandwidth.

 taa,
 Jedd.