On Tue, Jan 11, 2005 at 01:27:40PM -0600, Matthew A. Nicholson wrote:
What I am suggesting is the names internally stay the same
(com.name.branch.subbranch), but a new selector could be added that does
this:
name.com/branch/subbranch
or com.name.branch --> name.com/branch.
I like the idea of switching to host/subtree as the naming scheme, it
does seem much nicer. But doing it by creating multiple distinct
representations seems like a bad idea to me. I'd much rather just
remove the current impediments to using "/" in branch names, and then
switch over.
I'd actually at that point like to see Monotone _enforce_ the
requirement of having something that looks like <host>/<branch> for
branch names; experience says that even with the best intentions, if
we just rely on policy to make branch names unique, weird non-unique
things will escape and pollute the global namespace. It's just not
something that policy alone seems sufficient to enforce.
Of course, we wouldn't check that the <host> actually resolved, or try
and verify in 'whois' that the committer actually owned it, or
anything silly like that; but just a little syntactic requirement to
force people to think about the issue, and make the correct behavior
into the path of least resistance, would be good.
-- Nathaniel