[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: server down?
From: |
Francisco Vila |
Subject: |
Re: server down? |
Date: |
Fri, 3 Dec 2010 20:19:06 +0100 |
I am aware of the differences from our current system, but they are
not so big. Please continue reading:
2010/12/3 Colin Campbell <address@hidden>:
> On Thu, 2010-12-02 at 15:24 +0100, Francisco Vila wrote:
>> Git is already distributed and all the eggs are not in one basket. Wit
>> git installed in your system, you have a repository server just like
>> that in GNU. Gittorrent just removes restrictions on a centralised
>> management, which is more a political problem rather than technical.
>> LilyPond, like most projects, does have a central code repository and
>> a core development team: if we change that, we are talking about
>> another very different project.
>>
>
> I see your point, Francisco, but the difference is this: we have a
> single server which is considered canonical;
It is considered canonical because we consider it canonical. If it
were fully distributed, we still had to choose one as the canonical
one.
> you and I have *copies* of
> the repo,
Not exact. Git are not files we just copy. Git already consists of
multiple servers, one for each local repo which you call "a copy".
> and we can do as we like to our local version.
I consider a good thing that I can do whatever I like to my local version.
> Others have no
> knowledge of the state of our repo.
Others should have knowledge of my repo?
> When that single point of failure,
> the savannah server, goes offline, so does our definitive version of
> lilypond.
It is only definitive because we choose it to be. That is our is
policy, not a technical problem. If savannah becomes corrupted and I
trust on another repo more than in savannah's, what should be
considered definitive will move from savannah to that repo.
> The trust comes from access to a single physical version of
> the source.
So called "copies" are _equivalent_ physical versions. All commits
are signed, trust comes from the sha1 encryption.
> In a truly distributed, not just widely copied, VCS, the trust is based,
> for example, on credentials: only certain people can make changes,
Only certain people can already make changes, and that does not
prevent me from making changes on my local "copy".
> using
> some irrefutable signature.
Things already work that way.
> The changes are propagated in a torrent,
_That_ makes a difference.
> so
> the repository exists in its canonical form on many machines,
_that_ does not make a difference. It's already so.
> any one of
> which can be unavailable without compromising the whole.
The whole is already not compromised even if the central "copy" goes
offline. Just we don't have a permanent server for others constantly
seeing our repos from outside world. But when we push, there are no
differences between our repo and the remote. That is why I say it's
more a matter of policy than technical. No centralised version,
everybody sees others' repos, that is another project as I previously
said.
--
Francisco Vila. Badajoz (Spain)
www.paconet.org , www.csmbadajoz.com