> As far as various filesystems standards for what can and cannot be a
> filename,
> Let's raise the bar.  (Aren't filesytems supposed to maintain 
> the file's
> creation time
> themselves) (Not to mention maintaining the file's NAME).
> 
> Using timestamps for unique id's is a hack.  In a system like 
> Smalltalk
> where development
> cycle times can be under a second, this can lead to problems.  In any
> case it is just
> WRONG.  Still, it is usefull.  Dont even get me started on using
> Squeak's minute-based timestamps.

ChangeSet FILES are for distribution/interchange of Smalltalk code.
The current implementation uses a timestamp as part of the name in
order to provide UNIQUE names for multiple versions of change sets.
Other mechanisms could have been used, sequential numbers, random
numbers, with different implementation costs.

Another desirable trait is that common host tools easily display
ChangeSet files sorted first by name and then by creation date.
Appropriate choice for the timestamp supports both of these goals.

So far, the only real objection I've seen to using the ISO format is
that it consumes about 15 characters of the 32 character Mac file
name limit.  Add in the dots and the cs and you are left with only
13 characters for your ChangeSet name.

Perhaps an alternative implementation is to just number them sequentially.
There is already a time stamp in the change set for the change set and
on each method, and one in the host file system.

I would be happy to see:

myChangeSet.1
myChangeSet.2
myChangeSet.3

in my directories.

-david