On Monday, October 7, 2002, at 12:41 PM, Karl Fogel wrote:
> Colin Putney <cputney@whistler.net> writes:
>> The rule still applies. If auth info is specified on the command line,
>> use that info to authenticate. If authentication fails, the command
>> fails with an appropriate error. Otherwise, prompt.
>
> No, I'm talking about when the user *doesn't* specify the info on the
> command line.
Yes, I understand. That would fall under "otherwise, prompt."
>> Ah, this is a problem. The "explicit-auth" rule only works if the
>> prompting is necessary because of failed authentication. A script
>> shouldn't have to supply auth information for commands that don't
>> touch the repository. In that case, a --non-interactive flag is,
>> indeed, necessary.
>
> What I mean is, there are some commands that touch the repository, but
> which you can invoke *without* providing explicit auth info on the
> command line (instead, it's derived from the working copy -- or maybe
> it's absent entirely, because formerly the repository did not require
> auth and now suddenly it does).
Again, the rule still applies, and the use case falls under
"otherwise." If the user invokes a command without providing auth
information explicitly, and if authentication fails for whatever
reason, prompt for new auth info.
The rule is simple enough to be unambiguous for all cases involving
authentication. It falls down when there's a need to interact with the
user that is not related to authentication. I don't know what those
cases might be, but it doesn't seem appropriated to link their
behaviour to the presence of authentication command-line options.
Colin Putney
Whistler.com
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Oct 7 23:59:08 2002