Well this goes much too far, so this will be my last reply as this is going to 
be as off-topic as I can think of :)


----- Ursprüngliche Nachricht -----
Von: Grant Edwards
Gesendet am: 29 Sep 2009 19:16:18

On 2009-09-29, JMGross <[email protected]> wrote:

>> That's what I meant with a 'mesh cloud'. It still needs to be
>> stored somewhere. On a single place (that is hopefully
>> available) or on every single system in a copy, constantly
>> updating with the others (who are hopefully online and have a
>> recent version available).

>Except that's not how a distributed control system works.  It's
>not the case where there are mutliple mirrored copies of a
>single repository.  SVN uses a single repository on a single
>server, and there is no mechanism to allow the repository
>stored in multiple places.

No, everybody has a piece of the repository, or not, and you must be lucky to 
have at least someone online who has the piece you need. You cannot ever be 
sure that you'll get the latest revision (the two people with 
the latest are just offline) and you'll never notice.
It works better with an increasing base of users but there remains a 
statistical uncertainity.
The main problem is the same: if the information is not available, it is not 
available. With a dedicated server, no information or all is available, on a 
distributed system more or less is available (from nothing to all) and 
you'll never really know how much, unless there is again a central point where 
at least the metainformation is gathered.

>> If it's not stored somewhere, it is not available, no matter
>> which type of distribution is used. In my opinion, a dedicated
>> server (with some backups) is by far better than a repository
>> that is or is not available and is or is not up to date.

>I don't think you understand how distributed version control
>systems work.  It's not the same as having "a repository"
>that's stored in the cloud.

There are several different approaches. From a distributed repository (similar 
to the torrent DHC mechanism)  to a emule-like search in the cloud for the 
latest changes (with the same uncertainity).
I admittedly don't know how GIT or Bazaar exactly work, but as someone very 
clever once noticed: you cannot put more than a bit into a bit.
And even the mentioned wikipedia article (which does not tell much about the 
concepts) has in the 'disadvantages' section some points which are of big 
importance (even if desperately hidden between many 
'companies fear' and other pseudo-disadvantages)

>> The fact that TortoiseSVN (and I guesst SVN itself does so
>> too) keeps a working base together with the (actually altered)
>> working copy makes it a somewhat distributed repository too,

>Yup, that's how SVN works. You can do a diff/revert using just
>the local copy but most all of the other operations still
>require a live connection to the server.  DVCS systems don't.

So you want to tell me that I can revert to a previous version or compare two 
random versions of a file without either having all the history stored locally 
or a network connection? Well, fine piece of Magic. Information 
out of nowhere. If this is possible, I wonder why there are still programmers 
needed? It's all there, out of the void.

Each time, someone tells me he has developed a perpetuum mobile, a motor that 
requires less energy input as it outputs, or that he has pulled himself out of 
a swamp by its own hairs, I feel a strong urge do disbelief 
him.

So unless someone can tell me how this shall work (and not just that it works 
and I have to believe it), all types of DVCS I can imagine have as many 
drawbacks as they might have advantages.

>> Unfortunately there is no Linux version of Tortoise (only for
>> SVN itself), but since both are compatible and linux users (at
>> least the programmers) are usually fond of console usage... :)

>Tortoise doesn't really have anything to do with how SVN works,

And was just mentioned as an example of ease of use.
I already stated that it is linked together with SVN and of course is the 
repository job done by SVN.
But it makes version control as easy as right-clicking the mouse in the windows 
explorer (or any other app that uses the explorer view)

> There are plenty of other GUI interfaces to SVN, but
> there are still fundamental differences in the way than a
> single-reposity system works and the way that a distributed
> system works.

I never questioned that.
But I questioned that a DVCS system MUST be better than SVN (or other 
server-based systems like Perforce, which I also used many years ago or CVS 
which I refused to use even more years ago) just because it is 
a newer technology or more cool or whatever.
Both may have their strengths and both definitely have their drawbacks.
It depends on what you want to do with it and what features you need and what 
drawbacks you cannot afford.

JMGross


Reply via email to