Re: netrc field encryption in auth-source

[Robert Pluim]

Ted Zlatanov <address@hidden> writes:

> On Fri, 17 Jun 2011 09:17:13 +0200 Robert Pluim <address@hidden> wrote: 
> RP> Hmm, a single global setting works for me, but I can envisage people
> RP> desiring to have different values for different servers, not just
> RP> different files. At that point we'd be stuffing server values in the
> RP> custom variables, and we're back at square one. 
>
> Not quite square one, but yes, I know what you mean.  Per file regex is
> the lowest granularity I would implement without bribery, because it's
> what I would use.  But wait, we can do better if it's an alist...  Let's
> use the EPA file pattern!  The default can then be:
>
> `((,(car epa-file-auto-mode-alist-entry) nil)
>   (t ask))
>
> ...and when the user says "yes, use GPG tokens for file xyz" we'd add
> '("xyz" gpg) to the head of the alist and offer to save the defcustom.
> We have to make the "never ask to add" choice 'never, because nil is now
> a valid alist for the value.  So it could only be 'never or a valid
> alist.  Yes, that would work.
>

That sounds perfect to me.

> RP> Would a new keyword in the file itself work?  save-token
> RP> {cleartext,ask,gpg,no}? Tell me if I'm over-engineering this :)
>
> Every line in a netrc file should be self-sufficient, so I'd rather not
> add global keywords.  If you mean on each line, then yes, that's too
> much.

I did mean on each line, but it does seem like overkill. We can always
tell people to split things up into separate files and set preferences
per file.

Robert