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

[Andrea Arcangeli]

On Sat, Sep 27, 2003 at 12:03:52PM -0700, Tom Lord wrote:
> 
>     > From: Andrea Arcangeli <address@hidden>
> 
>     > They're definitely not data.
> 
>     > The only reason they exists is to track down the evolution of the
>     > project, 
> 
> No, they're also useful for whole-tree changesets.  Another use might
> be source auditing.

the very auditing you have to do, is to check that no tagline is
replicated in two files or that two people added the same tagline at the
same time, something worthless, that 'arch' can't guarantee.

arch can generates unique tagids by itself without having to spend cpu
time at all to audit for it, you only have to let arch handle the metadata
transparently instead of asking the end user to generate it.

The user should never see with his eyes a cryptic meaningless (for him) tag-ID 
in
his whole life, the same way he doesn't need to see the contents of the
patchset tarball, and arch can provide a mathematically safe behaviour
with my-id and the sequence number. As I outlined in point 3 there will
be no chance of collisions, total reliable behaviour. Unlike the tagline
mode that works probabilistically (unless you want really really ugly
tags like address@hidden where I keep the sequence counter count by
hand, again a repetitive work that arch should do not me and it'd be
still more reliable to let the software do it).

> You could think of it not as "tracking" code, although it can be used
> for such, but also as a tool for making changesets possible.

there is no need of such a tool, arch can do changesets just fine
without adding this "tool" to the data.

>     > The reason this is obviously not data, is that an user downloading
>     > linux-2.4.22.tar.gz pure data package, will have absolutely zero
>     > benefits from the tagline embedded in the source, the tagline for this
>     > use will be pure _pollution_. 
> 
> You can repeat that as many times as you like, of course.

I will until I will convice you ;). More seriously, this is clearly a
minor issue, I mean, arch just looks great and I could live just fine by
only making sure the explicit method works better and safer than the
tagline (the my-id+ sequence number or any equivalent math safe
approch). All I care about is that the software fits my needs, if others
prefer to use it in a different way that's their own right, but I'm
quite convinced the tagline method shouldn't be "recommended".

And personally I'd rather add a metadata channel to the inode than to
use taglines as they're right now.

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/