Public SourceMage Discussion List

["Jeremy Blosser (emrys)" <jblosser-smgl AT firinn.org>]

On Aug 19, Eric Sandall [eric AT sandall.us] wrote:
> On Fri, 19 Aug 2005, Seth Alan Woolley wrote:
> > On Fri, Aug 19, 2005 at 07:42:35PM +0200, Ladislav Hagara wrote:
> >>>
> >>> be a problem. I would prefer four or five people who had their public
> >>> keys in a keyring instead of one commonly shared key everybody has
> >>> access to -- then we can trace guilt ;).
> >>>
> >> Personally I don't like big changes in perforce and moreover I hate
> >> binary files there.
> >> It is very difficult to check it out.
> >>
> >> What about if in perforce were stored only individual keys and the
> >> keyring was created on the user's box as part of scribe update?
> 
> The individual keys would still be binaries...so we'd now have lots of
> little binaries instead of one big one?

We could store the individual keys as ascii versions.  I'm not sure how
that would help with whatever the issues are with binaries; it's just an
ascii encoding, so it's not like text diffs are going to make sense between
revisions.  If anything it's worse since the encoded ascii version is
larger than the binary version.

> > Then if you got a keyring, you'd have to know how to break it apart into
> > separate files and our specific format -- as long as this process was
> > documented, I think it would be ok.  Perhaps we can store ascii armored
> > public keys with the --minimal* gpg switches as tag-name.pgpk files
> > which then are coalesced into tag.gpg files?
> >
> > We should make them part of tarball creation, too, not another thing
> > done on scribe update, or does it matter to anybody if we decentralize
> > this part?
> 
> I'd rather keep this simple. Our current setup seems to work and I see
> no problems with it, whereas this would add more complexity to `scribe
> update` (or wherever we would decide to put the repackaging of
> tag.gpg).

I agree, but if there are other benefits to keeping these ascii that I'm
not seeing, we could just keep the full keyrings as ascii in the grimoires
(p4 and on user machines) and make the verificatin routines do the
conversion to a binary keyring as needed.  I'm not sure how bad that
performance hit would be so there may still be a need to cache if we did
that.

Attachment: pgpVM9j7sdiKa.pgp
Description: PGP signature