issue #898

[Edmund Wong <ed...@kdtc.net>]

Hi,

I'm reading up on the issues and came across #898 which
talks about rename = copy+delete (paraphrased summary).

Perhaps I've misunderstood the whole issue, and please
correct me if I'm wrong; but, what if rename = move?
If I understand the issue correctly, copy + delete requires
two 'activities' whereas move is a single activity which
dispenses with doing two 'atomic' commits. (After all, a copy
  action is logged and a commit action is logged.)  Perhaps a
modified move?   Then might this genereate IDs that are
different (and won't conflict with future actions)?

Sometimes I should learn more about patience, in my
earnest to contribute without foreknowledge in the
subject.  This young grasshopper's cup is empty.

Edmund

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=1559698

Re: issue #898

["C. Michael Pilato" <cm...@collab.net>]

Branko Cibej wrote:
> Greg Stein wrote:
>> On Mon, Apr 6, 2009 at 10:23, Branko Cibej <br...@xbc.nu> wrote:
>>   
>>> ...
>>> At one point we even had a true rename in the filesystem, but it turned
>>> out to be sort of useless because the client and RA layers and editors
>>> couldn't handle it.
>>>     
>> Then you may be happy to hear that wc-ng will have the capability to
>> record moves. I've also got move operations in my editor redesign
>> thinking. Of course, layers that don't understand that will see
>> copy/delete, but we can start to propagate move information as far as
>> possible.
>>   
> 
> Yup, saw that, and indeed I am happy, happy! :)

Based on my experience with the FS atomic renames design and implementation,
I remain cautiously pessimistic about how far these changes will take us.
That doesn't discount the goodness of the changes, though, of course.

-- 
C. Michael Pilato <cm...@collab.net>
CollabNet   <>   www.collab.net   <>   Distributed Development On Demand

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=1562829

Re: issue #898

[Branko Cibej <br...@xbc.nu>]

Greg Stein wrote:
> On Mon, Apr 6, 2009 at 10:23, Branko Cibej <br...@xbc.nu> wrote:
>   
>> ...
>> At one point we even had a true rename in the filesystem, but it turned
>> out to be sort of useless because the client and RA layers and editors
>> couldn't handle it.
>>     
>
> Then you may be happy to hear that wc-ng will have the capability to
> record moves. I've also got move operations in my editor redesign
> thinking. Of course, layers that don't understand that will see
> copy/delete, but we can start to propagate move information as far as
> possible.
>   

Yup, saw that, and indeed I am happy, happy! :)

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=1560175

Re: issue #898

[Greg Stein <gs...@gmail.com>]

On Mon, Apr 6, 2009 at 10:23, Branko Cibej <br...@xbc.nu> wrote:
>...
> At one point we even had a true rename in the filesystem, but it turned
> out to be sort of useless because the client and RA layers and editors
> couldn't handle it.

Then you may be happy to hear that wc-ng will have the capability to
record moves. I've also got move operations in my editor redesign
thinking. Of course, layers that don't understand that will see
copy/delete, but we can start to propagate move information as far as
possible.

Cheers,
-g

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=1560020

Re: issue #898

[Branko Cibej <br...@xbc.nu>]

Edmund Wong wrote:
> Hi,
>
> I'm reading up on the issues and came across #898 which
> talks about rename = copy+delete (paraphrased summary).
>
> Perhaps I've misunderstood the whole issue, and please
> correct me if I'm wrong; but, what if rename = move?
> If I understand the issue correctly, copy + delete requires
> two 'activities' whereas move is a single activity which
> dispenses with doing two 'atomic' commits. (After all, a copy
>   action is logged and a commit action is logged.)  Perhaps a
> modified move?   Then might this genereate IDs that are
> different (and won't conflict with future actions)?
>
> Sometimes I should learn more about patience, in my
> earnest to contribute without foreknowledge in the
> subject.  This young grasshopper's cup is empty.
>   

Yes, one of my pet gripes about our implementation of rename is that it
does not preserve "object identity", which means that you have to jump
through hoops to tell the difference between a branch, a copy, and a rename.

At one point we even had a true rename in the filesystem, but it turned
out to be sort of useless because the client and RA layers and editors
couldn't handle it.

This rename thing is not the only case of "let's make it look like it
works first" in Subversion. :)

-- Brane

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=1559773