Opened 8 years ago
Last modified 8 years ago
#368 new bug
even mo better approach for filesync — at Initial Version
Reported by: | Fred T. Hamster | Owned by: | |
---|---|---|---|
Priority: | minor | Milestone: | |
Component: | feistymeow-octopus | Version: | |
Keywords: | Cc: |
Description
keep a file in the dir that's the md5sumish result!
so we don't have to keep computing them all the time!
for this, we should be able to trust the dates!
the timestamp on the file tells us if we need to redo the summary
file.
also summary file's non-existence tells us to create one! wheee!!!!
but this rocks!
it gets rid of a lot of the problems of the current approach without
a lot of changes.
however;
the summary info would be our wacky strided checksums right?
well, if the file size hasn't changed then we can still do it right.
so the summary file, if it exists,
should encode the:
file time (even though summary file timestamp should be later)
size of file
info about file
if timestamp or size changed, redo the summary!
oh everyone's buddha this is good.
would fix so much about the file synch that maybe the other shit wouldn't matter?
if we use dot files, at least those things are hidden.
hey, what about grouping all our records per dir!
we already have file timestamp separate!
also totally tasty! then max of one file per dir, and it's
easy to scan the dir to see if things changed.
do we care if people don't want a single dot file added per dir
on both sides?
NO, not at this point. maybe later.
oooh ooooh!
the kicker here!
use json format for the summary file.
NICE!