Re: [Sks-devel] [PATCH] Bundle IPv6, DNS fixes, sks dump fix

[Phil Pennock]

On 2009-03-25 at 09:36 +0000, Kim Minh Kaplan wrote:
> I have finally bundled recent changes into a single patch against Yaron
> Minsky's repository.  It includes sks dump fix[1], DNS staleness fix[2]
> and IPv6 support as Phil Pennock suggested it[3] but with additional
> rewritings:

You dropped the fix for not complaining of malformed lines when a line
has just a comment in it.

You've dropped the clean-up of the address stringification, which puts
square brackets around addresses; addresses with colons followed by
:port is confusing, especially then there's a :: address shortening in
there somewhere.

You've dropped the reload_mailsync_if_changed fix to ensure that the
mtime of !Settings.mailsync_file is compared to the stored mtime of
!mailsync_partners instead of !membership, which will fix a case of
reloading-mailsync-every-time.

sks.pod was patched to provide a better sksconf example, removing the
deprecated logfile item and setting initial_stat (takes < 5 seconds on
my box, which is beginning to show its age) and disable_ipv6.  I noted
in a follow-up that apparently the only working syntax for setting bools
is "boolname:" because otherwise !Settings.anonlist gets littered and
commands like dump get extraneous parameters from the config.  In light
of this, and previous posters having trouble with "sks dump" using
strange file-names, I do think it's really important to get in a boolean
option in the example sksconf as there's a natural inclination to
provide a value on/true/yes/1.  "initial_stat:" is IMO a good example to
keep here.

On the bright side, these are all *really* simple fixes.  :)

For the new config rules of yours, sks.pod should probably show how to
enable IPv6 cleanly with:
  hkp_address: 0.0.0.0 ::
to just bind to multiple families -- assuming of course that this is
supported (it represents the defaults in my version).  I'm barely awake
to respond, so may very well be wrong, but when I double-check, I think
that Common.match_client_recon_addr needs adjusting to deal with this.
Might be as simple as just special-casing INADDR_ANY / in6addr_any.

Regards,
-Phil

pgpzoOvHMKxQJ.pgp
Description: PGP signature