Re: [Monotone-devel] Re: how to merge trees

[jc007b6093]

This is why I think directories should contain some form of UID for the
files in them not just the file name.

Joel

> On Tue, Aug 16, 2005 at 01:31:23PM +0100, Bruce Stephens wrote:
>> Nathaniel Smith <address@hidden> writes:
>> >   -- once files are dead, they stay dead.  If some revision R is a
>> >      descendent of B, and R does not contain the file, that file will
>> >      not be contained in any descendent of R.
>>
>> Doesn't the second of those make deletion rather too sticky?
>> Specifically I'm imagining forking at R (or somewhere later), and in
>> one branch the file gets deleted, but in the other it doesn't.
>>
>> If I understand everything correctly, I must choose to kill the file
>> in a subsequent merge (since the merge is a descendant of both
>> branches).  But surely that can't be so, so what am I
>> misunderstanding?
>
> It does make deletion rather too sticky.  You're not misunderstanding.
>
> (You can, of course, create a new file with the same name, in the
> merged version.)
>
> OTOH, it is the same model we've always had, and this version of it
> actually works :-).  (I'm coming to respect "works" a _lot_ when it
> comes to merge algorithms; it turns out that, at least as of
> recently, we just had, like... none that did.)  So, baby steps...
> "resurrection" is a non-trivial thing to support correctly.  For the
> reasons you mention, I'd like to support it anyway... but I don't want
> to until I can be confident we have a good representation that can be
> handled safely.
>
> -- Nathaniel
>
> --
> "...All of this suggests that if we wished to find a modern-day model
> for British and American speech of the late eighteenth century, we could
> probably do no better than Yosemite Sam."
>
>
> _______________________________________________
> Monotone-devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/monotone-devel
>
>