You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Greg Stein <gs...@lyra.org> on 2000/06/06 10:34:13 UTC
Re: [subversion-dev] Children knowing about their parents
On Tue, 6 Jun 2000, Manoj Kasichainula wrote:
> The statement that children need to know about all their parents seems
> to have some bad implications to me.
I can't find this reference. ??
> If a child has n names, then it
> naturally needs to point to n parent nodes. But doesn't a file have to
> point to every version of its parent node that has the file as its
> child? And everytime you update a directory, you need to update all
> children of that directory to point to the new parent. This seems
> unscalable to me. Am I missing something obvious here?
The issue has come up, and I had thought it was last left as "there are no
parent pointers." Part of the reason for that is to prevent the need to
alter historical file revisions (to include the new parent).
I can see that scaling would also be a bit of a bummer.
AFAIK, we construct a new directory and point it at new/old children,
bubbling that on up. The old children are untouched; they just happen to
be reachable via a new path.
Cheers,
-g
--
Greg Stein, http://www.lyra.org/
Re: [subversion-dev] Children knowing about their parents
Posted by Karl Fogel <kf...@galois.collab.net>.
Manoj Kasichainula <ma...@collab.net> writes:
> 6.1.3.2 Bubble-Up Method, second-to-last paragraph before first
> diagram:
>
> "Parent-child links go both ways (i.e., a child knows who all its
> parents are), but a node's name is stored only in its parent, because
> a node with multiple parents may have different names in different
> parents."
>
> > The issue has come up, and I had thought it was last left as "there are no
> > parent pointers." Part of the reason for that is to prevent the need to
> > alter historical file revisions (to include the new parent).
>
> Maybe this was just a left over paragraph that wasn't supposed to
> remain?
That's exactly what it was, actually. We're just going to
double-check that it is completely false, and then remove it if so.
JimB, Ben, and I all have some other priority work today, however, so
the offending paragraph will remain in a bit longer... :-)
Thanks for catching this one, though!
-K
Re: [subversion-dev] Children knowing about their parents
Posted by Manoj Kasichainula <ma...@collab.net>.
On Tue, Jun 06, 2000 at 03:34:13AM -0700, Greg Stein wrote:
> On Tue, 6 Jun 2000, Manoj Kasichainula wrote:
> > The statement that children need to know about all their parents seems
> > to have some bad implications to me.
>
> I can't find this reference. ??
6.1.3.2 Bubble-Up Method, second-to-last paragraph before first
diagram:
"Parent-child links go both ways (i.e., a child knows who all its
parents are), but a node's name is stored only in its parent, because
a node with multiple parents may have different names in different
parents."
> The issue has come up, and I had thought it was last left as "there are no
> parent pointers." Part of the reason for that is to prevent the need to
> alter historical file revisions (to include the new parent).
Maybe this was just a left over paragraph that wasn't supposed to
remain?