Dropping non-machine readable

David Shaw <[email protected]> Sat, 16 Apr 2005 09:54:39 -0400
Newsgroups gmane.comp.encryption.pgp.keyserver-folk
Message-ID <[email protected]>
Hi folks,

When I first added the --search-keys feature to GnuPG, searching an
HKP server was done by reading the "index" response and parsing the
HTML for key data.  This was, and is, very messy and prone to failure,
as any change to the HTML output breaks things (which happened a
number of times).  To resolve this, back in 2002 I added the "MR"
machine readable output to PKSD, then Yaron Minsky added it to SKS,
and Jonathan McDowell added it to ONAK.  There may be other HKP-like
servers that support it as well.

I am working on removing this HTML parsing code, as it significantly
complicates the HKP handler and makes some other potential changes
(libcurl, for one) more difficult.  I only intended to keep the HTML
parser until MR was adopted and that seems to have happened.

A quick glance at the servers in the subkeys.pgp.net rotation shows
that all but one of the servers currently support MR.  Same thing for
the wwwkeys.pgp.net rotation (the same server doesn't support MR in
both rotations).  pgp.mit.edu supports MR.

Note that some HKP servers don't support MR and also don't return data
in the PKSD HTML index form.  I'm not particularly concerned about
these servers - they didn't work with GnuPG searching before, and
still won't work after this change, so no functionality has been lost.

To give some transition time, I will not make this change final before
October of this year, and even after the change is made, I will
include the old code for a while so it that can be enabled at build
time.

David