On 9/7/07, Ben Collins-Sussman <sussman@red-bean.com> wrote:
> On 9/7/07, Erik Huelsmann <ehuels@gmail.com> wrote:
>
> > Why is a python script better than adding it as /experimental/ to the
> > CLI? I think we may need to amend the versioning policy to do that
> > though.
> >
> > Why I think this is acceptable: the CLI isn't a C API...
> >
>
> I think that CLI compatibility guarantees are *more* restrictive that
> API guarantees. Most users don't see the APIs at all, yet they
> absolutely habituate to CLI UIs.
The reason we have (in the past) regarded the CLI as an API (at least
as voiced) is that scripts can call out to the command line and
changing options and commands would break those scripts. (People can
re-train...)
> We've made small CLI tweaks here and there (added switches), but we've
> never done something as radical and removing a whole subcommand. I
> think that's why people are afraid to add "experimental" CLI commands.
Why would the above be different when we add an external script?
People will habituate to that too... Yet the script will - eventually
- get scrapped too.
bye,
Erik.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri Sep 7 19:11:21 2007