Let's say that I'm running Subversion on webdav.org and hosting projects
there. I send you these URLs so that you can check out stuff:
Do you care where the repository starts? I think not. You know that you can
checkout the URLs that I just gave you.
As administrator of the machine, I organize the projects in a way that makes
the most sense for my machine.
There is no simple syntactic marker that you can place into a URL to
separate the pieces, and still retain the ability to surf through them (for
example, say we put the repos path after a "?" mark; that would blow away
surfing thru the repos). If we imposed a specific string (a path component)
to occur within the URL, then we're impacting how a person designs and
presents their server.
Consider my www.lyra.org web site. I want to put the whole bugger under
version control. I don't want users to surf my site using URLs like:
I want them to use http://www.lyra.org/greg/index.html.
Where a repository begins within the URL space is a private issue to the
server. It shouldn't be exposed to users. You give them a base URL for a
checkout, and they know it works. What do they care if the parent is or is
not under version control?
The use-case listed below asks about checking out a larger parent context. I
don't see the problem there. They try to check out
http://svn.webdav.org/projects/ and it works or it fails. They can or they
can't. Knowing the repository boundary isn't going to change that.
Can you tell me why we *should* have a boundary? To what purpose? What
feature does it enable that we can't have otherwise?
On Fri, Feb 16, 2001 at 12:56:59PM -0600, Ben Collins-Sussman wrote:
> OK, there's an unresolved issue between Greg S. and Karl: whether
> a repository path and fs path should be exposed as seperate objects to
> the user (as cvs does), or if they should be expressed as a single URL
> (as dav would do it).
> I've not yet heard people put out real arguments one way or another;
> and I'd like to hear them *now*. I'm tired of beating around the
> Here's what I've heard from Greg:
> * `Con' on separation:
> Separation is an old holdover from cvs; the user doesn't need
> to be aware of it.
> Here's what I've heard from Karl:
> * `Con' on urls:
> Use-case: Jane checks out a url, then realizes that she wants
> to checkout a larger parent context. But which parent dir? She
> can't tell where the repos starts! Needless obscurity.
> Obviously, because I sit next no Karl, I hear his side more. However,
> I'm sure that Greg has specific pros/cons... please post them!
> -- Referee Ben
Greg Stein, http://www.lyra.org/
Received on Sat Oct 21 14:36:22 2006