Is this right?

[Russel Winder]

Vincent,

At the risk of extending a debate beyond its natural life, doing things
at too detailed a level, and appearing to be too much of a nit-picker...

On Tue, 2008-11-04 at 09:17 +0100, Vincent Ladeuil wrote:
> >>>>> "Russel" == Russel Winder <russel.winder at concertant.com> writes:

> 'screw' seems a bit excessive here, the branch is not updated so
> it's as good as it was before the pull (merge, missing ?).

The word "screw" has many, many meanings both as a noun and a verb.  I
think you are taking a stronger meaning than the colloquial English (at
least London) meaning.

>     Russel> everyone using a branch of this branch
> 
> And this is good. That branch has ceased to exist. You got a copy
> the last time you pulled from it, that's the best you can hope in
> a distributed world.

That is true but not the point.  The point is about notice and
notification.  We must be careful about the potential to make a mountain
out of a molehill -- i.e. we should not do it.  Despite my grumpiness
this is overall a small issue.

> We can blame Guillermo, but what if, instead of changing his
> name, he created another lp user ?
> 
> You get no notice that the branch you're pointing to is now a
> dead one and will forever stay subscribed.

I'm afraid we had that situation previously as well. :-(

> And what if you're subscribed to a branch that must be relocated
> on another site because (put any reason here) ?
> 
> For any change outside of your control, you have to act.

My email was really about courtesy not functionality, and I think we
must not lose sight of that.  The issue here is about dissemination of
information not about bzr functionality.

>     Russel> since there appears to have been no announcement
>     Russel> anywhere of the change.
> 
> I don't think announcements are a way to solve that *kind* of
> problem. 
> 
> In that specific case it could have helped you, yes, but in
> general there is no link between a branch and its subscribers, so
> there is now way to warn all the subscribers.

Exactly my point.  There is a set of unknown and unnamed people that
deserve an announcement that things are about to break, or have just
been broken.  I am not at all worried that the change happens, I just
think that announcing that the change is about to happen, or even that
it has already happened, is a simple courtesy.

>     >> Rusell, running: bzr pull --remember lp:bzr-xmloutput  should fix it.
> 
>     Russel> It does appear to have done.  But remember everyone
>     Russel> has the bzr+ssh: protocol URL stored so another
>     Russel> change will cause people further aggravation.
> 
> We could blame the way 'lp:' protocol works here as it chases the
> pointer and gives back
> bzr+ssh://bazaar.launchpad.net/%7Everterok/bzr-xmloutput/trunk/
> which indeed is another failure waiting to happen.
> 
> Or we could blame verterok for not putting his branch under
> bzr+ssh://bazaar.launchpad.net/~bzr-xmloutpu/bzr-xmloutput/trunk/
> which should remain valid longer.
> 
> But in the end, branch URLs just change, that's a fact of life.

lp: works the way it works.  It might be better that it were the lp:
address that bzr stores but that is not the way things work.  Choosing
to use a project-based rather than person-based URL for a plugin that is
integral rather than peripheral to the bzr milieu might be a good move,
but that is not really the point.  Nor is the fact that Guillermo has
chose to change the name of the branch actually a serious issue.  There
are undoubtedly good reasons why he wanted to do this.  None of this is
a problem per se.

>     Russel> I think failing to announce this sort of change
> 
> Even if we had the list of the branches based on this branch,
> what can be done ? Part of these are obviously not reachable,
> some may have cease to exist, that list should also include a way
> to contact the branch owner, what about anonymous owners ?
> 
> Etc, etc.

The technology is not at issue here.  The issue is the social structure
not the technical structure.

>     Russel> and assuming that everyone will divine that it has
>     Russel> happened 
> 
> You got a BzrNotABranch error, which is the best bzr can do
> without a crystal ball (it can't divine either).

Indeed.

>     Russel> and will know exactly what to do, is not a way of
>     Russel> progressing.
> 
> In that specific case, you can argue that lp could redirect the
> old ID branches to the new ones (but that means keeping the old
> ID in use) and in that case, depending on how lp implement it,
> bzr may even warn you that a redirection has occurred... but
> that's really a tiny fraction of a branch death possible causes.

Actually I don't see this technological point as either needed nor the
right response to the situation.  I would prefer to keep the technology
simple -- VCS is complex enough without bending it to deal with dark
corner situations where a technological solution is not actually the
answer.

>     Russel> I should note that this is not the first time...
> 
> And certainly not the last. I could also mention having been
> instructed to subscribe to a branch which was not updated as
> regularly as the one I really wanted...

Indeed.

>     Russel> Apologies for being grumpy on this 
> 
> This is understandable, I went into it a couple of times too.

I guess this one just came at a moment where I felt I had to email
rather than just "turn a blind eye".

What I do want to say though is that the technology we are discussing
here seems to Just Work (tm) so the efforts Guillermo and others put in
to making it work are well focused -- Bazaar works in Eclipse after all.
We must no forget in all these irritations that Bazaar works in Eclipse
and other IDEs.  Whilst I am an Emacs and command line user myself,
Eclipse, NetBeans and IntelliJ IDEA are the front line of Bazaar take up
-- along with TortoiseSVN, Olive-GTK and QBzr.  The code works, lets not
forget that.

>     Russel> but it is these small things that get to be
>     Russel> the most irritating.
> 
> This is certainly not a small thing but a social or
> organizational problem, nothing bzr can address other than
> providing the easiest way to take the change into account when
> the new url is provided. And this is exactly what --remember
> does.
> 
> But if you have any idea on how to handle the general problem
> better, I'm surely interested.

To be honest all that is needed is a small announcement email on the
email list and a note on the Lauchpad pages.  The issue is public
information presented in places where people can find it easily.

When I came across the problem, I went to the xmloutput page on
Launchpad -- nothing obvious.  Neither is it obvious from the fact that
lp protocol URL is unchanged that there is a problem.  True if you dig
deeper then the fact appear.  The problem is that people are prone not
to dig deeper first, they do a surface investigation and then if the
fail, they either complain (as I did publicly) or they quietly go away
and then ignore the technology.  This last is the worst case scenario.

-- 
Russel.
====================================================
Dr Russel Winder                 Partner

Concertant LLP                   t: +44 20 7585 2200, +44 20 7193 9203
41 Buckmaster Road,              f: +44 8700 516 084
London SW11 1EN, UK.             m: +44 7770 465 077
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20081104/d51d6921/attachment.pgp