Re: [Gnu-arch-users] Re: give us a hand with arch

[Andrea Arcangeli]

On Sat, Sep 27, 2003 at 11:26:23AM -0700, David Brown wrote:
> On Sat, Sep 27, 2003 at 09:31:04AM -0700, Tom Lord wrote:
> 
> > Another point of view says that the idea of an inventory id is a
> > usefully universal one that transcends the issues of revision control.
> 
> What if we made the format of the tags more "universal", something that
> could conceivably by used by systems other than arch.
>
> Perhaps:
> 
>  <punct> file-id: "text" <punct>
> 
> or even something that makes it stand out more.  That might help make it
> clear that the file tag isn't really an arch thing, it is just that arch
> is the only thing that can do anything useful with it.
> 
> Other systems make use of unique id's for files or other pieces of data.
> Uuids get fairly common use these days.  I don't know if we want to
> restrict the id's to a 64 bit integer, though.

it would be still obviously metadata, something completely worthless and
pure pollution for users fetching the tarball and not caring about the
revision history, nor to track down moves of those files when inside a
system like arch or bitkeeper. There is no point in keeping this
metadata mixed with the data. Infact you even mathematically risks
collisions when you add new files that way, and you would also need to
check that this tag isn't replicate twice in the whole tree which is an
overhead with thousnad of files. You pollute the namespace of a language
as well.

Who does really care not having to execute add-tag/delete-tag/move-tag?
In a big project you want strict commits anyways.  Once those functions
and the explicit mode will be high powered (or at least when the only
disavantage will be having to call add-tag/delete-tag/move-tag), there
will be no issue any longer with them and the metadata will be fine to
stay separate from the data, and it'll be more reliable too by using a
sequence number and my-id as I suggested in point 3 in a past email.
Then I think the tagline and names could be obsoleted as well.

I already made the point that to be universal and to work with all sorts
of languages and binary formats, you *can't* mix data with metadata. The
default mode should be changed to be explicit, then if you want to leave
tagline for backwards compatibility or for smaller projects that's
certainly fine with me.

Andrea - If you prefer relying on open source software, check these links:
            rsync.kernel.org::pub/scm/linux/kernel/bkcvs/linux-2.[45]/
            http://www.cobite.com/cvsps/