W3C home > Mailing lists > Public > w3c-dist-auth@w3.org > January to March 2007

Re: binds and overwriting infinite-depth copies

From: Geoffrey M Clemm <geoffrey.clemm@us.ibm.com>
Date: Wed, 17 Jan 2007 16:00:55 -0500
To: "Tim Olsen" <tolsen718@gmail.com>
Cc: w3c-dist-auth@w3.org
Message-ID: <OFCFFF576F.E4C97DE2-ON85257266.00738C28-85257266.0073D0C5@us.ibm.com>
Your option (4) is fine with me.  Now that I think of it, this did come up 
a while ago, and at that time, I think we expressed a preference for 
options 1/2 (keep the destination structure, and "last copy wins").

But in my opinion, one could make a reasonable argument for any of the 4 
choices (so I'd code up whatever is easiest, unless you know for some 
reason that your clients would prefer one of those options).

Cheers,
Geoff
 



"Tim Olsen" <tolsen718@gmail.com> 
01/17/2007 02:57 PM

To
Geoffrey M Clemm/Lexington/IBM@IBMUS
cc
w3c-dist-auth@w3.org
Subject
Re: binds and overwriting infinite-depth copies






If I go with option 3 and R3 is a VCR, then R4 and R5 will not contain
the same version history any longer.  (See answer to this thread:
http://lists.w3.org/Archives/Public/ietf-dav-versioning/2006JulSep/0002.html
)

Would this be a possible option?

/b/c/p - R3
/b/d/p - R4
where R3 is a copy of R1
where R4 is a copy of R2

This way the structure is similar to /a and one of the resources
carries on as R3

thanks,
Tim

On 1/17/07, Geoffrey M Clemm <geoffrey.clemm@us.ibm.com> wrote:
>
> It probably is easy to figure out, but below, I meant:
>   "is more structurally equivalent to /a"
> and
>   "it is reasonable for the server to do 1 or 2".
>
> Sorry for the poor proofreading on the original response!
>
> Cheers,
> Geoff
>
>
>  geoff wrote on 01/17/2007 07:05:05 AM:
> >
>  > The spec does not give a definitive answer to this one.  You have
>  > three reasonable choices:
>  >
>  > /b/c/p - R3
>  > /b/d/p - R3
>  > where R3 is a copy of R1
>  >
>  > /b/c/p - R3
>  > /b/d/p - R3
>  > where R3 is a copy of R2
>  >
>  > /b/c/p - R4
>  > /b/d/p - R5
>  > where R4 is a copy of R1
>  > where R5 is a copy of R2
>  >
>  > I have a slight preference for the third choice, since it is
>  > symmetric and is more structurally equivalent to /b.
>  > But I think it is reasonable for a server to do 1 or 3, in case it
>  > is expensive for it to detect this situation (so I think the spec
>  > should leave this up to the server).
>
>  > "Tim Olsen" <tolsen718@gmail.com>
>  > Consider the following case.  There exist the following URLs and the
>  > resource's they are bound to:
>  >
>  > /a/c/p - R1
>  > /a/d/p - R2
>  > /b/c/p - R3
>  > /b/d/p - R3
>  >
>  > /b/c and /b/d are different collections.
>  >
>  > What should happen if I do a COPY /a /b  with overwrite set to true?
>  > Should the new /b/c/p and /b/d/p still be the same resource?  Keep in
>  > mind that R3 may be a VCR.
>
Received on Wednesday, 17 January 2007 21:01:00 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 2 June 2009 18:44:15 GMT