Re: git tag's history

From: Nicolas Sebrecht <nicolas.s.dev@gmx.fr>
Date: Jul 26, 2010, 14:17 UTC
Message-ID: <20100726141755.GA13721@vidovic>
In-Reply-To: <1280151367.8229.28.camel@wpalmer.simply-domain>

The 26/07/10, Will Palmer wrote:

Show 5 quoted lines

> On Mon, 2010-07-26 at 02:36 -0700, David Aguilar wrote:
> > He won't be able to push the tag unless he forcibly deletes your tag first.
> > 
> > Tags do not have history.  This is a non-technical problem.  If s/he's
> > deleting your tags, it's a social problem.

True in the first place but I see nothing wrong to want to have technical safeguards around social issues.

Show 5 quoted lines

> What about enabling the reflog on the shared bare repository?
> Tags get changed. That's a fact. It's a necessity, no matter what the
> theoretical model says. Sometimes there's a reason to replace a tag, and
> when that happens there is sometimes a reason to hang on to the old
> value.

Sharing a central repository makes it exposed on such erasure. Git provides other mechanisms to take more control. Your workflow may not be what everybody want for themselves. You should look at the hooks system, I think.

You might also be interested by gitolite, an external tool (I'm not sure it could help on this particular issue, though).

>        Pretending "it's a social issue" ignores reality, not the least
> important aspect of which is "social issues happen. How do we avoid
> losing information in a sane, scalable manner?"

It shouldn't be lost. The bare repository can be seen as a gate to publicize local work (or a backup copy sometimes). The original tag can be restored by his creator from his local repository.

> Just an example of a tag which could stand to be updated, take a look
> at: refs/tags/junio-gpg-pub  There's at least one thing in there which
> is out of date.

Hard tags are designed to be immutable whatever happens after the facts. This is why they exist.

Public tags shouldn't change as it would break other up-to-date repositories.

-- 
Nicolas Sebrecht