[PATCH consfigurator 3/3] introduce the pass data store.

[Sean Whitton]

Hello,

On Sun 20 Feb 2022 at 11:11pm -04, David Bremner wrote:

> Sean Whitton <spwhitton at spwhitton.name> writes:
>
>> I'm wary of letting HOST not be a hostname.  It breaks with the
>> documented convention, and might clash with possible extensions of this
>> data source.  How about for extensibility safety, we validate that HOST
>> is an actual hostname and otherwise return nil?
>
> My reasoning was that people (i.e. me) already have pass(1) stores, and
> I expect most of them to be somewhat hierarchical (based on e.g. the
> pass docs).  If we don't allow things like "--user-passwd-irc/freenode"
> then we're basically telling people they have to reorganize them.

I see what you mean.  Maybe you could probably just add symlinks to your
repo, so it's some extra work but not a reorganisation that could
interfere with other usage?

> One option is to not insist on the prefix, and only insist on
> hostname looking HOST values (via regex?) for appropriately prefixed
> IDEN1.

Do you mean: whenever the prefix is --user-passwd- then it must be
followed by an actual hostname, but otherwise IDEN1 can be anything and
is not transformed?  It's less extensible, but maybe it would mean other
extensions are not required.  What I was suggesting was exactly that but
other IDEN1 are rejected.

Regex is what I had in mind for validation, yes.

Let me know what sounds best to you atm.

-- 
Sean Whitton