Re: [Sks-devel] another bounds problem in SKS

[Yaron Minsky]

Presentation filter is only used for presentation to end users.  The
underlying replication is all exactly the same.  So, they will fetch 
the original, unfiltered key.

y


On Mon, 4 Oct 2004 00:34:54 +0200, Peter Palfrader <address@hidden> wrote:
> On Sun, 03 Oct 2004, Yaron Minsky wrote:
> 
> 
> 
> > On Sun, 3 Oct 2004 10:58:17 -0400, Yaron Minsky <address@hidden> wrote:
> > > On Sun, 3 Oct 2004 15:29:09 +0200, Dinko Korunic <address@hidden> wrote:
> > > >
> > > > Is this normal? Are all of these keys having the wrong MPI size?
> > > >
> > > > 2004-10-03 15:26:51 /pks/lookup: Index request: (0xc5c0c5c7218d18d7)
> > > > 2004-10-03 15:26:54 Error fetching uid during VIndex for keyid 
> > > > 0x6C76178CA0ED982D: ParsePGP.Overlong_mpi
> > > > 2004-10-03 15:26:54 Error fetching uid during VIndex for keyid 
> > > > 0xBF41595B7372A2A5: ParsePGP.Overlong_mpi
> > > > 2004-10-03 15:26:56 Error fetching uid during VIndex for keyid 
> > > > 0x6C76178CA0ED982D: ParsePGP.Overlong_mpi
> > > > 2004-10-03 15:26:56 Error fetching uid during VIndex for keyid 
> > > > 0xBF41595B7372A2A5: ParsePGP.Overlong_mpi
> > > > 2004-10-03 15:26:57 Error fetching uid during VIndex for keyid 
> > > > 0x6C76178CA0ED982D: ParsePGP.Overlong_mpi
> > > > 2004-10-03 15:26:58 Error fetching uid during VIndex for keyid 
> > > > 0xBF41595B7372A2A5: ParsePGP.Overlong_mpi
> > > >
> >
> > First, yes, those keys are busted.
> >
> > Second, you can now check for yourself.  With the latest patch
> > (patch-22), you can ask for the raw key without any presentation
> > filters required.  Putting "clean=off" in the request url will disable
> > all key cleaning. So now you can see the status of those keys by
> > yourself, by grabbing the  key directly and looking at it via pgpdump.
> >
> > Third, those error messages won't come up anymore, since keys are now
> > by default cleaned before the UID is fetched from them.
> >
> > y
> >
> > (note, if no further patches come up, I will tag this as 1.0.8 and
> > release it by Monday.)
> 
> So, if two servers recon, they will exchange hashes of the "broken"
> keys.  Then, they will fetch the key from the peer, using a get on the
> hash.
> 
> Will they fetch a key with or without presentation filters applied?
> 
> --
> Peter
> 
> 
> 
> 
> _______________________________________________
> Sks-devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/sks-devel
>