I doesn't break MC itself, but the proper fix for external packages,
Gofer, Metacello, is to improve their code to operate on the new
first-class MCVersionNames; e.g., _they_ should send #versionName, if
they don't want the extension or ancestry info.

VersionNames are now a first-class object, the idea is to get away
from the same String processing (copyUpTo: blah blah) sprinkled
everywhere.

 - Chris


On Tue, Mar 22, 2011 at 12:12 AM,  <commits at source.squeak.org> wrote:
> A new version of Monticello was added to project The Inbox:
> http://source.squeak.org/inbox/Monticello-ul.442.mcz
>
> ==================== Summary ====================
>
> Name: Monticello-ul.442
> Author: ul
> Time: 20 March 2011, 5:30:01.846 am
> UUID: 14eda109-f1ba-4940-adb0-c48470081d21
> Ancestors: Monticello-ul.441
>
> - reverted MCFileBasedRepository >> #allVersionNames, because it breaks quite a lot of stuff (like Gofer, Metacello, probably MC itself). The cause of the problem is that the   "file extensions" and previous versions in case of .mcds are not stripped from the result. Existing code concatenates the requested "file extension" to the end resulting in .mcz.mcz http requests which obviously fail.
>
> =============== Diff against Monticello-ul.441 ===============
>
> Item was changed:
>  ----- Method: MCFileBasedRepository>>allVersionNames (in category 'private-files') -----
>  allVersionNames
> +       ^ self readableFileNames collect: [:ea | (ea copyUpToLast: $.) copyUpTo: $(]!
> -       ^ self readableFileNames!
>
>
>