[darcs-users] [planet darcs] Automatic detection of file renames for Darcs
AntC
anthony_clayden at clear.net.nz
Thu Aug 22 06:16:44 UTC 2013
> José Neder <jlneder <at> gmail.com> writes:
> ...
Thanks José for the quick response.
> I don't know if any other vcs use it but i guess not because i didn't
> see a feature like this anywhere else.
Me neither. I wonder why?
> I didn't read all the paper but i guess you means the labels used to
> uniquely identify files? It is mentioned in the subsection 4.1 Rename
> Files.
Yes. And in section 4.2 Directory moves.
> You could say is a unique identifier but one assigned by the
> filesystem and not the repo, so it is useful for the almost the same
> things mentioned in the paper.
I think the paper is going one step further (section 5.2 Moving patches
across repos; and Section 6 comparison to darcs 'adapting representation'
meaning that "patches cannot be signed" -- although the discussion is
rather sketchy). I think that what they're aiming for is a file id that is
globally unique across all repos.
> Putting it shortly inode are not persistent between filesystems ...
> For any other filesystem or machine the inode number means nothing, ...
Yes, I didn't expect inode to be unique/persistent globally. That's why I
suggested prefixing inode with the repo id (plus machine id?) to arrive at
a globally unique file id. (I know there's a regular discussion topic in
darcsland about GUID's.)
Then wherever the file gets pulled/pushed, the patch could carry the
file's GUID. Each repo would maintain a map of file GUID <-> inode-in-my-
repo.
> ... [inodes] are almost like any other number you can choose at random.
Sounds exactly like a GUID to me ;-)
>
> 2013/8/22 AntC <anthony_clayden <at> clear.net.nz>:
> > I see that José Neder is doing some work to use inode ...
> >
> > Hi José,
> > are inode and Windows' File Index number guaranteed persistent through
> > renames/directory moves and edits of the file?
> >
> > It seems an easy way to keep track of files. Do other VCS's use it? Is
> > there some reason darcs hasn't used it before?
> >
> > inode sounds like in the 'Principled Approach to Version Control'
paper:
> > unique identifiers internal to the repo.
> > If DARCS could prefix inode with a unique repo identifier, that would
give
> > a GUID for every file wherever it gets pulled(?)
> >
More information about the darcs-users
mailing list