While waiting for a file to save, all of a sudden it struck me:
Save images as composites! Although there is a lot of metadata in a PSD, the layers can easily be separated, and why should we resave one, wasting CPU/disk I/O and time, when it hasn't changed? Apologies if you already realise this and have implemented it, but I want to go a step further.
Say we have a large file made up of 100 evenly spaced tiles. Then small edits to one of them does not require the whole to be resaved, saving up to 99% save time... Similarly, if only half change, we can still save a lot of time! They may not even need to be evenly spaced... Obviously then they can be combined in a compressed archive.
Not sure if Ps Dev team like the idea... can someone 'tag them' or how do I make them see it?
Well, but you still need to hold the pixels for preview in memory (and save it, when using maximum compatibility) and depending on how the actual internal data structures are organized it may in fact be even more resource consuming to look up the relevant data sections/ chunks and splice them out. Also you'd still need integrity tests to verify the data is saved and handled okay. That and of course that you could at least in part already mimic a similar workflow using smart objects. Still, doesn't really eliminate any of the previous. File I/O issues are really solved best by brute force - large transfer bandwidth, fast disks.