On Jun 21, 2007, at 0:49 , Lex Spoon wrote:

Bert Freudenberg <[EMAIL PROTECTED]> writes:
On Jun 20, 2007, at 23:31 , Lex Spoon wrote:
Maybe it is not even too late. Could the tools simply detect that the
packages are using generated uuid's, and then guess a "uuid" by
looking at the filename?  E.g., for a file named "Etoys-edc.23",
quietly treating the uuid as "Etoys"?

Huh? Fo merging you *really* need the most recent common ancestor,
and you need to be *sure* which it is. UUIDs work well for this.
*Anything* could have happened between the two versions that were
saved under the same name and you would completely lose these changes
when comparing names only.

Maybe I have the terms wrong, but I still do not see how a UUID helps
with this.  You would identify a specific ancestor by name ("Etoys")
plus version ("edc.23").  What is the problem?

The problem is that this is not unique. Like, the actual problem we were discussing is that Edgar accidentally saves two versions with the same name ("edc.23").

Going the other way, do you truly want to have a repository that has
two separate packages named Etoys-edc.23?  That sounds like a recipe
for chaos.

Right, one should not do this. But that's just policy, or common sense if you will.

MC itself does not put any meanings into version names. You cannot tell form a version name which one is newer, or to what package it belongs. It's just a human-readable name.

Now, the squeaksource server UI as well as the MC UI tools actually interpret the names according to established conventions. This makes live easier if you adopt these rules, too (and harder if you ignore them). They should still work, none-theless, if you give arbitrary names to versions.

Given the distributed development model behind MC UUIDs are about the only choice. For example, it's well possible two developers use the same initials. Both produce version 23. Then what?


- Bert -


_______________________________________________
V3dot10 mailing list
[email protected]
http://lists.squeakfoundation.org/mailman/listinfo/v3dot10

Reply via email to