Re: [Gnu-arch-users] tla import versus tla commit

[Robert Collins]

On Wed, 2004-10-13 at 18:06 -0500, Matthew Dempsky wrote:
> Cameron Patrick <address@hidden> writes:
> 
> > Anand Kumria wrote:
> >
> >> [1]: tla import having -S and -s is a bad idea; perhaps -C,
> >> --create-archive would be less likely to confuse. I've typed in '-S' more
> >> than once.
> >
> > --create-archive would be misleading as the option doesn't create an
> > archive, just creates the category/branch/version within it.  I'd say
> > that there's a good case to be made for -S being the default on tag
> > and especially import, though.
> 
> Making -S the default would be convenient, but keep in mind the
> possibility of typoing a category, branch, or version number.
> Creating a new directory and sticking some contents in it mean that
> there's no possibility of reusing the name in the future.
> 
> If you had typoed archive-setup you would create an empty directory
> which is safe to destroy, and a typoed tag/import command would alert
> you that the directory doesn't exist.

Of the number if tags issues, what percent are to existing (populated or
not) branches? 1% ? 0.5 % ?

Assume 1%. 99% of the time, -S will be added by the user. It becomes
muscle memory. And once that happens, it may as well be the default
because few if any folk will be -ever- using archive-setup + tag without
-S ... remember - optimise for the common case, handle the exception
gracefully.

In this case, the exception is that someone may need to remove the
tagged branch completely - and users routinely do that in #arch. In
fact, they tend to remove the entire archive, and only do that
partially. I'm starting to think we should have a 'nuke a tagged or
impotred version' facility, where we get the user to confirm that its
not been mirrored etc, and which only operates on versions with 0 or 1
revisions.

Rob

signature.asc
Description: This is a digitally signed message part